1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // Test host codegen.
3 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK1
4 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
5 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK2
6 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK3
7 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
8 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK4
9 
10 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
12 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
14 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
15 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
16 
17 // Test target codegen - host bc file has to be created first.
18 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
19 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
20 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
21 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK10
22 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
23 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
24 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
25 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK12
26 
27 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
28 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
29 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
30 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
31 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
32 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
33 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
34 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
35 
36 // Test host codegen.
37 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK17
38 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
39 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK18
40 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK19
41 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
42 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK20
43 
44 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
45 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
46 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
47 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
48 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
49 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
50 
51 // Test target codegen - host bc file has to be created first.
52 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
53 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK25
54 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
55 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK26
56 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
57 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK27
58 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
59 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK28
60 
61 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
62 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
63 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o %t %s
64 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
65 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
66 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
67 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -o %t %s
68 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -fopenmp-is-device -fopenmp-host-ir-file-path %t-x86-host.bc -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
69 
70 // expected-no-diagnostics
71 #ifndef HEADER
72 #define HEADER
73 
74 
75 
76 
77 // We have 8 target regions, but only 7 that actually will generate offloading
78 // code, only 6 will have mapped arguments, and only 4 have all-constant map
79 // sizes.
80 
81 
82 
83 // Check target registration is registered as a Ctor.
84 
85 
86 template<typename tx, typename ty>
87 struct TT{
88   tx X;
89   ty Y;
90 };
91 
92 int global;
93 
94 int foo(int n) {
95   int a = 0;
96   short aa = 0;
97   float b[10];
98   float bn[n];
99   double c[5][10];
100   double cn[5][n];
101   TT<long long, char> d;
102 
103   #pragma omp target teams distribute num_teams(a) thread_limit(a) firstprivate(aa) nowait
104   for (int i = 0; i < 10; ++i) {
105   }
106 
107   #pragma omp target teams distribute if(target: 0)
108   for (int i = 0; i < 10; ++i) {
109     a += 1;
110   }
111 
112 
113   #pragma omp target teams distribute if(target: 1)
114   for (int i = 0; i < 10; ++i) {
115     aa += 1;
116   }
117 
118 
119 
120   #pragma omp target teams distribute if(target: n>10)
121   for (int i = 0; i < 10; ++i) {
122     a += 1;
123     aa += 1;
124   }
125 
126   // We capture 3 VLA sizes in this target region
127 
128 
129 
130 
131 
132   // The names below are not necessarily consistent with the names used for the
133   // addresses above as some are repeated.
134 
135 
136 
137 
138 
139 
140 
141 
142 
143 
144 
145   #pragma omp target teams distribute if(target: n>20) dist_schedule(static, n)
146   for (int i = 0; i < 10; ++i) {
147     a += 1;
148     b[2] += 1.0;
149     bn[3] += 1.0;
150     c[1][2] += 1.0;
151     cn[1][3] += 1.0;
152     d.X += 1;
153     d.Y += 1;
154   }
155 
156   return a;
157 }
158 
159 // Check that the offloading functions are emitted and that the arguments are
160 // correct and loaded correctly for the target regions in foo().
161 
162 
163 
164 
165 // Create stack storage and store argument in there.
166 
167 // Create stack storage and store argument in there.
168 
169 // Create stack storage and store argument in there.
170 
171 // Create local storage for each capture.
172 
173 
174 
175 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
176 
177 template<typename tx>
178 tx ftemplate(int n) {
179   tx a = 0;
180   short aa = 0;
181   tx b[10];
182 
183   #pragma omp target teams distribute if(target: n>40)
184   for (int i = 0; i < 10; ++i) {
185     a += 1;
186     aa += 1;
187     b[2] += 1;
188   }
189 
190   return a;
191 }
192 
193 static
194 int fstatic(int n) {
195   int a = 0;
196   short aa = 0;
197   char aaa = 0;
198   int b[10];
199 
200   #pragma omp target teams distribute if(target: n>50)
201   for (int i = a; i < n; ++i) {
202     a += 1;
203     aa += 1;
204     aaa += 1;
205     b[2] += 1;
206   }
207 
208   return a;
209 }
210 
211 struct S1 {
212   double a;
213 
214   int r1(int n){
215     int b = n+1;
216     short int c[2][n];
217 
218     #pragma omp target teams distribute if(target: n>60)
219     for (int i = 0; i < 10; ++i) {
220       this->a = (double)b + 1.5;
221       c[1][1] = ++a;
222     }
223 
224     return c[1][1] + (int)b;
225   }
226 };
227 
228 int bar(int n){
229   int a = 0;
230 
231   a += foo(n);
232 
233   S1 S;
234   a += S.r1(n);
235 
236   a += fstatic(n);
237 
238   a += ftemplate<int>(n);
239 
240   return a;
241 }
242 
243 
244 
245 // We capture 2 VLA sizes in this target region
246 
247 
248 // The names below are not necessarily consistent with the names used for the
249 // addresses above as some are repeated.
250 
251 
252 
253 
254 
255 
256 
257 
258 
259 
260 
261 
262 
263 
264 
265 
266 
267 
268 
269 
270 
271 // Check that the offloading functions are emitted and that the arguments are
272 // correct and loaded correctly for the target regions of the callees of bar().
273 
274 // Create local storage for each capture.
275 // Store captures in the context.
276 
277 
278 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
279 
280 
281 // Create local storage for each capture.
282 // Store captures in the context.
283 
284 
285 
286 
287 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
288 
289 // Create local storage for each capture.
290 // Store captures in the context.
291 
292 
293 
294 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
295 
296 #endif
297 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooi
298 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
299 // CHECK1-NEXT:  entry:
300 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
301 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
302 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
303 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
304 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
305 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
306 // CHECK1-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
307 // CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
308 // CHECK1-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
309 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
310 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
311 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
312 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
313 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
314 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
315 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
316 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
317 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
318 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
319 // CHECK1-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
320 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
321 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
322 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
323 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
324 // CHECK1-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
325 // CHECK1-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
326 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
327 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
328 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
329 // CHECK1-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
330 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_22:%.*]] = alloca i32, align 4
331 // CHECK1-NEXT:    [[A_CASTED23:%.*]] = alloca i64, align 8
332 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED25:%.*]] = alloca i64, align 8
333 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [10 x i8*], align 8
334 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [10 x i8*], align 8
335 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [10 x i8*], align 8
336 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
337 // CHECK1-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
338 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
339 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
340 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
341 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
342 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
343 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
344 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
345 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
346 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
347 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
348 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
349 // CHECK1-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
350 // CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
351 // CHECK1-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
352 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
353 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
354 // CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
355 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
356 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
357 // CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
358 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
359 // CHECK1-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
360 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
361 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
362 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
363 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
364 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
365 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
366 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
367 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
368 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
369 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
370 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
371 // CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
372 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
373 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
374 // CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
375 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
376 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
377 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
378 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
379 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
380 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
381 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
382 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
383 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
384 // CHECK1-NEXT:    store i8* null, i8** [[TMP24]], align 8
385 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
386 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
387 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
388 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
389 // CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
390 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
391 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
392 // CHECK1-NEXT:    store i8* null, i8** [[TMP29]], align 8
393 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
394 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
395 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
396 // CHECK1-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
397 // CHECK1-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
398 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
399 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
400 // CHECK1-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
401 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
402 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
403 // CHECK1-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
404 // CHECK1-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 120, i64 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
405 // CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
406 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
407 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
408 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
409 // CHECK1-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
410 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
411 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
412 // CHECK1-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
413 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
414 // CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
415 // CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
416 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
417 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
418 // CHECK1-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
419 // CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
420 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
421 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
422 // CHECK1-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
423 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP53]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
424 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
425 // CHECK1-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
426 // CHECK1-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
427 // CHECK1-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP38]])
428 // CHECK1-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
429 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
430 // CHECK1-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
431 // CHECK1-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
432 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
433 // CHECK1-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
434 // CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
435 // CHECK1-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
436 // CHECK1-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
437 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
438 // CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
439 // CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
440 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
441 // CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
442 // CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
443 // CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
444 // CHECK1-NEXT:    store i8* null, i8** [[TMP65]], align 8
445 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
446 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
447 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
448 // CHECK1-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
449 // CHECK1-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
450 // CHECK1-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
451 // CHECK1:       omp_offload.failed:
452 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i64 [[TMP60]]) #[[ATTR3]]
453 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
454 // CHECK1:       omp_offload.cont:
455 // CHECK1-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
456 // CHECK1-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
457 // CHECK1-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
458 // CHECK1-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
459 // CHECK1-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
460 // CHECK1-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
461 // CHECK1-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
462 // CHECK1-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
463 // CHECK1-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
464 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
465 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
466 // CHECK1:       omp_if.then:
467 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
468 // CHECK1-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
469 // CHECK1-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
470 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
471 // CHECK1-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
472 // CHECK1-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
473 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
474 // CHECK1-NEXT:    store i8* null, i8** [[TMP79]], align 8
475 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
476 // CHECK1-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
477 // CHECK1-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
478 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
479 // CHECK1-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
480 // CHECK1-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
481 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
482 // CHECK1-NEXT:    store i8* null, i8** [[TMP84]], align 8
483 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
484 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
485 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
486 // CHECK1-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP85]], i8** [[TMP86]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
487 // CHECK1-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
488 // CHECK1-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
489 // CHECK1:       omp_offload.failed20:
490 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
491 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
492 // CHECK1:       omp_offload.cont21:
493 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
494 // CHECK1:       omp_if.else:
495 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
496 // CHECK1-NEXT:    br label [[OMP_IF_END]]
497 // CHECK1:       omp_if.end:
498 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
499 // CHECK1-NEXT:    store i32 [[TMP89]], i32* [[DOTCAPTURE_EXPR_22]], align 4
500 // CHECK1-NEXT:    [[TMP90:%.*]] = load i32, i32* [[A]], align 4
501 // CHECK1-NEXT:    [[CONV24:%.*]] = bitcast i64* [[A_CASTED23]] to i32*
502 // CHECK1-NEXT:    store i32 [[TMP90]], i32* [[CONV24]], align 4
503 // CHECK1-NEXT:    [[TMP91:%.*]] = load i64, i64* [[A_CASTED23]], align 8
504 // CHECK1-NEXT:    [[TMP92:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_22]], align 4
505 // CHECK1-NEXT:    [[CONV26:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED25]] to i32*
506 // CHECK1-NEXT:    store i32 [[TMP92]], i32* [[CONV26]], align 4
507 // CHECK1-NEXT:    [[TMP93:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED25]], align 8
508 // CHECK1-NEXT:    [[TMP94:%.*]] = load i32, i32* [[N_ADDR]], align 4
509 // CHECK1-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[TMP94]], 20
510 // CHECK1-NEXT:    br i1 [[CMP27]], label [[OMP_IF_THEN28:%.*]], label [[OMP_IF_ELSE35:%.*]]
511 // CHECK1:       omp_if.then28:
512 // CHECK1-NEXT:    [[TMP95:%.*]] = mul nuw i64 [[TMP2]], 4
513 // CHECK1-NEXT:    [[TMP96:%.*]] = mul nuw i64 5, [[TMP5]]
514 // CHECK1-NEXT:    [[TMP97:%.*]] = mul nuw i64 [[TMP96]], 8
515 // CHECK1-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
516 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP98]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i64 80, i1 false)
517 // CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
518 // CHECK1-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i64*
519 // CHECK1-NEXT:    store i64 [[TMP91]], i64* [[TMP100]], align 8
520 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
521 // CHECK1-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i64*
522 // CHECK1-NEXT:    store i64 [[TMP91]], i64* [[TMP102]], align 8
523 // CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
524 // CHECK1-NEXT:    store i8* null, i8** [[TMP103]], align 8
525 // CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
526 // CHECK1-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
527 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 8
528 // CHECK1-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
529 // CHECK1-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
530 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 8
531 // CHECK1-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 1
532 // CHECK1-NEXT:    store i8* null, i8** [[TMP108]], align 8
533 // CHECK1-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 2
534 // CHECK1-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i64*
535 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP110]], align 8
536 // CHECK1-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 2
537 // CHECK1-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i64*
538 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP112]], align 8
539 // CHECK1-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 2
540 // CHECK1-NEXT:    store i8* null, i8** [[TMP113]], align 8
541 // CHECK1-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 3
542 // CHECK1-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
543 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP115]], align 8
544 // CHECK1-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 3
545 // CHECK1-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
546 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP117]], align 8
547 // CHECK1-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
548 // CHECK1-NEXT:    store i64 [[TMP95]], i64* [[TMP118]], align 8
549 // CHECK1-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 3
550 // CHECK1-NEXT:    store i8* null, i8** [[TMP119]], align 8
551 // CHECK1-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 4
552 // CHECK1-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
553 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 8
554 // CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 4
555 // CHECK1-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
556 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 8
557 // CHECK1-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 4
558 // CHECK1-NEXT:    store i8* null, i8** [[TMP124]], align 8
559 // CHECK1-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 5
560 // CHECK1-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i64*
561 // CHECK1-NEXT:    store i64 5, i64* [[TMP126]], align 8
562 // CHECK1-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 5
563 // CHECK1-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
564 // CHECK1-NEXT:    store i64 5, i64* [[TMP128]], align 8
565 // CHECK1-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 5
566 // CHECK1-NEXT:    store i8* null, i8** [[TMP129]], align 8
567 // CHECK1-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 6
568 // CHECK1-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
569 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP131]], align 8
570 // CHECK1-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 6
571 // CHECK1-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i64*
572 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP133]], align 8
573 // CHECK1-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 6
574 // CHECK1-NEXT:    store i8* null, i8** [[TMP134]], align 8
575 // CHECK1-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 7
576 // CHECK1-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
577 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 8
578 // CHECK1-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 7
579 // CHECK1-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
580 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 8
581 // CHECK1-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
582 // CHECK1-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 8
583 // CHECK1-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 7
584 // CHECK1-NEXT:    store i8* null, i8** [[TMP140]], align 8
585 // CHECK1-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 8
586 // CHECK1-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
587 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 8
588 // CHECK1-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 8
589 // CHECK1-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
590 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 8
591 // CHECK1-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 8
592 // CHECK1-NEXT:    store i8* null, i8** [[TMP145]], align 8
593 // CHECK1-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 9
594 // CHECK1-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i64*
595 // CHECK1-NEXT:    store i64 [[TMP93]], i64* [[TMP147]], align 8
596 // CHECK1-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 9
597 // CHECK1-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
598 // CHECK1-NEXT:    store i64 [[TMP93]], i64* [[TMP149]], align 8
599 // CHECK1-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 9
600 // CHECK1-NEXT:    store i8* null, i8** [[TMP150]], align 8
601 // CHECK1-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
602 // CHECK1-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
603 // CHECK1-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
604 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
605 // CHECK1-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
606 // CHECK1-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
607 // CHECK1-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED33:%.*]], label [[OMP_OFFLOAD_CONT34:%.*]]
608 // CHECK1:       omp_offload.failed33:
609 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
610 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT34]]
611 // CHECK1:       omp_offload.cont34:
612 // CHECK1-NEXT:    br label [[OMP_IF_END36:%.*]]
613 // CHECK1:       omp_if.else35:
614 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
615 // CHECK1-NEXT:    br label [[OMP_IF_END36]]
616 // CHECK1:       omp_if.end36:
617 // CHECK1-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
618 // CHECK1-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
619 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
620 // CHECK1-NEXT:    ret i32 [[TMP156]]
621 //
622 //
623 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
624 // CHECK1-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
625 // CHECK1-NEXT:  entry:
626 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
627 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
628 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
629 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
630 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
631 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
632 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
633 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
634 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
635 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
636 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
637 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
638 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
639 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
640 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
641 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
642 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
643 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
644 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
645 // CHECK1-NEXT:    ret void
646 //
647 //
648 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
649 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
650 // CHECK1-NEXT:  entry:
651 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
652 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
653 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
654 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
655 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
656 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
657 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
658 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
659 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
660 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
661 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
662 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
663 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
664 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
665 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
666 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
667 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
668 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
669 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
670 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
671 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
672 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
673 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
674 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
675 // CHECK1:       cond.true:
676 // CHECK1-NEXT:    br label [[COND_END:%.*]]
677 // CHECK1:       cond.false:
678 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
679 // CHECK1-NEXT:    br label [[COND_END]]
680 // CHECK1:       cond.end:
681 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
682 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
683 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
684 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
685 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
686 // CHECK1:       omp.inner.for.cond:
687 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
688 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
689 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
690 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
691 // CHECK1:       omp.inner.for.body:
692 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
693 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
694 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
695 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
696 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
697 // CHECK1:       omp.body.continue:
698 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
699 // CHECK1:       omp.inner.for.inc:
700 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
701 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
702 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
703 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
704 // CHECK1:       omp.inner.for.end:
705 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
706 // CHECK1:       omp.loop.exit:
707 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
708 // CHECK1-NEXT:    ret void
709 //
710 //
711 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
712 // CHECK1-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
713 // CHECK1-NEXT:  entry:
714 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
715 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
716 // CHECK1-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
717 // CHECK1-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
718 // CHECK1-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
719 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
720 // CHECK1-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
721 // CHECK1-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
722 // CHECK1-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
723 // CHECK1-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
724 // CHECK1-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
725 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
726 // CHECK1-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
727 // CHECK1-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
728 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
729 // CHECK1-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
730 // CHECK1-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
731 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
732 // CHECK1-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
733 // CHECK1-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
734 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
735 // CHECK1-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
736 // CHECK1-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
737 // CHECK1-NEXT:    ret void
738 //
739 //
740 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
741 // CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
742 // CHECK1-NEXT:  entry:
743 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
744 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
745 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
746 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
747 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
748 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
749 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
750 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
751 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
752 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
753 // CHECK1-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
754 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
755 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
756 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
757 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
758 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
759 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
760 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
761 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
762 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
763 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
764 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
765 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
766 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
767 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
768 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
769 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
770 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
771 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
772 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
773 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
774 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
775 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
776 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
777 // CHECK1-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
778 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
779 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
780 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
781 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
782 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
783 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
784 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
785 // CHECK1-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !21
786 // CHECK1-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !21
787 // CHECK1-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !21
788 // CHECK1-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !21
789 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
790 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
791 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
792 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
793 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
794 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
795 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
796 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
797 // CHECK1-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
798 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
799 // CHECK1-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
800 // CHECK1:       omp_offload.failed.i:
801 // CHECK1-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
802 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
803 // CHECK1-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !21
804 // CHECK1-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !21
805 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
806 // CHECK1-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
807 // CHECK1-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !21
808 // CHECK1-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !21
809 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
810 // CHECK1-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
811 // CHECK1-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !21
812 // CHECK1-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !21
813 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
814 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
815 // CHECK1:       .omp_outlined..1.exit:
816 // CHECK1-NEXT:    ret i32 0
817 //
818 //
819 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
820 // CHECK1-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
821 // CHECK1-NEXT:  entry:
822 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
823 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
824 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
825 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
826 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
827 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
828 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
829 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
830 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
831 // CHECK1-NEXT:    ret void
832 //
833 //
834 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
835 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
836 // CHECK1-NEXT:  entry:
837 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
838 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
839 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
840 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
841 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
842 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
843 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
844 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
845 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
846 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
847 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
848 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
849 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
850 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
851 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
852 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
853 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
854 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
855 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
856 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
857 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
858 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
859 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
860 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
861 // CHECK1:       cond.true:
862 // CHECK1-NEXT:    br label [[COND_END:%.*]]
863 // CHECK1:       cond.false:
864 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
865 // CHECK1-NEXT:    br label [[COND_END]]
866 // CHECK1:       cond.end:
867 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
868 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
869 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
870 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
871 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
872 // CHECK1:       omp.inner.for.cond:
873 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
874 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
875 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
876 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
877 // CHECK1:       omp.inner.for.body:
878 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
879 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
880 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
881 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
882 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
883 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
884 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[CONV]], align 4
885 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
886 // CHECK1:       omp.body.continue:
887 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
888 // CHECK1:       omp.inner.for.inc:
889 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
890 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
891 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
892 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
893 // CHECK1:       omp.inner.for.end:
894 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
895 // CHECK1:       omp.loop.exit:
896 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
897 // CHECK1-NEXT:    ret void
898 //
899 //
900 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
901 // CHECK1-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
902 // CHECK1-NEXT:  entry:
903 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
904 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
905 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
906 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
907 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
908 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
909 // CHECK1-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
910 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
911 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
912 // CHECK1-NEXT:    ret void
913 //
914 //
915 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
916 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
917 // CHECK1-NEXT:  entry:
918 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
919 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
920 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
921 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
922 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
923 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
924 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
925 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
926 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
927 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
928 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
929 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
930 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
931 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
932 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
933 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
934 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
935 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
936 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
937 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
938 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
939 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
940 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
941 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
942 // CHECK1:       cond.true:
943 // CHECK1-NEXT:    br label [[COND_END:%.*]]
944 // CHECK1:       cond.false:
945 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
946 // CHECK1-NEXT:    br label [[COND_END]]
947 // CHECK1:       cond.end:
948 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
949 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
950 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
951 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
952 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
953 // CHECK1:       omp.inner.for.cond:
954 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
955 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
956 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
957 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
958 // CHECK1:       omp.inner.for.body:
959 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
960 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
961 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
962 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
963 // CHECK1-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
964 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
965 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
966 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
967 // CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
968 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
969 // CHECK1:       omp.body.continue:
970 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
971 // CHECK1:       omp.inner.for.inc:
972 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
973 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
974 // CHECK1-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
975 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
976 // CHECK1:       omp.inner.for.end:
977 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
978 // CHECK1:       omp.loop.exit:
979 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
980 // CHECK1-NEXT:    ret void
981 //
982 //
983 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
984 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
985 // CHECK1-NEXT:  entry:
986 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
987 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
988 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
989 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
990 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
991 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
992 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
993 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
994 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
995 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
996 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
997 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
998 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
999 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1000 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
1001 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1002 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
1003 // CHECK1-NEXT:    ret void
1004 //
1005 //
1006 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
1007 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
1008 // CHECK1-NEXT:  entry:
1009 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1010 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1011 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1012 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1013 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1014 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1015 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1016 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1017 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1018 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1019 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1020 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1021 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1022 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1023 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1024 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1025 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1026 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1027 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
1028 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1029 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1030 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1031 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1032 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1033 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1034 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
1035 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1036 // CHECK1:       cond.true:
1037 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1038 // CHECK1:       cond.false:
1039 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1040 // CHECK1-NEXT:    br label [[COND_END]]
1041 // CHECK1:       cond.end:
1042 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1043 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1044 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1045 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1046 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1047 // CHECK1:       omp.inner.for.cond:
1048 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1049 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1050 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1051 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1052 // CHECK1:       omp.inner.for.body:
1053 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1054 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
1055 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1056 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1057 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
1058 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
1059 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
1060 // CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
1061 // CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
1062 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
1063 // CHECK1-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
1064 // CHECK1-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
1065 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1066 // CHECK1:       omp.body.continue:
1067 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1068 // CHECK1:       omp.inner.for.inc:
1069 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1070 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
1071 // CHECK1-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
1072 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1073 // CHECK1:       omp.inner.for.end:
1074 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1075 // CHECK1:       omp.loop.exit:
1076 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1077 // CHECK1-NEXT:    ret void
1078 //
1079 //
1080 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
1081 // CHECK1-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1082 // CHECK1-NEXT:  entry:
1083 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1084 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
1085 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1086 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
1087 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
1088 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1089 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
1090 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
1091 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
1092 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1093 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1094 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1095 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1096 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
1097 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1098 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
1099 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
1100 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1101 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
1102 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
1103 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
1104 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1105 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1106 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
1107 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1108 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
1109 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
1110 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1111 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
1112 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
1113 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
1114 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1115 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
1116 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1117 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
1118 // CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
1119 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
1120 // CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
1121 // CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
1122 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
1123 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
1124 // CHECK1-NEXT:    ret void
1125 //
1126 //
1127 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
1128 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1129 // CHECK1-NEXT:  entry:
1130 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1131 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1132 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1133 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
1134 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1135 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
1136 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
1137 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1138 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
1139 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
1140 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
1141 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1142 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1143 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1144 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1145 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1146 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1147 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1148 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1149 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1150 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1151 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1152 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
1153 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1154 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
1155 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
1156 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1157 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
1158 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
1159 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
1160 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1161 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1162 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
1163 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1164 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
1165 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
1166 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1167 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
1168 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
1169 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
1170 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1171 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1172 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
1173 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1174 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1175 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
1176 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1177 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1178 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
1179 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1180 // CHECK1:       omp.dispatch.cond:
1181 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1182 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
1183 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1184 // CHECK1:       cond.true:
1185 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1186 // CHECK1:       cond.false:
1187 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1188 // CHECK1-NEXT:    br label [[COND_END]]
1189 // CHECK1:       cond.end:
1190 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1191 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1192 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1193 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1194 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1195 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1196 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1197 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1198 // CHECK1:       omp.dispatch.body:
1199 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1200 // CHECK1:       omp.inner.for.cond:
1201 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1202 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
1203 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1204 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1205 // CHECK1:       omp.inner.for.body:
1206 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1207 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1208 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1209 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
1210 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !22
1211 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
1212 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !22
1213 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
1214 // CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !22
1215 // CHECK1-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
1216 // CHECK1-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
1217 // CHECK1-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
1218 // CHECK1-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !22
1219 // CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
1220 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
1221 // CHECK1-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
1222 // CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
1223 // CHECK1-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
1224 // CHECK1-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
1225 // CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
1226 // CHECK1-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
1227 // CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
1228 // CHECK1-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
1229 // CHECK1-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
1230 // CHECK1-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
1231 // CHECK1-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
1232 // CHECK1-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
1233 // CHECK1-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
1234 // CHECK1-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
1235 // CHECK1-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
1236 // CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
1237 // CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !22
1238 // CHECK1-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
1239 // CHECK1-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !22
1240 // CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
1241 // CHECK1-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !22
1242 // CHECK1-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
1243 // CHECK1-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
1244 // CHECK1-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
1245 // CHECK1-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !22
1246 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1247 // CHECK1:       omp.body.continue:
1248 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1249 // CHECK1:       omp.inner.for.inc:
1250 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1251 // CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
1252 // CHECK1-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1253 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
1254 // CHECK1:       omp.inner.for.end:
1255 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1256 // CHECK1:       omp.dispatch.inc:
1257 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1258 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1259 // CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
1260 // CHECK1-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
1261 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1262 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1263 // CHECK1-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
1264 // CHECK1-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
1265 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
1266 // CHECK1:       omp.dispatch.end:
1267 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
1268 // CHECK1-NEXT:    ret void
1269 //
1270 //
1271 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
1272 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1273 // CHECK1-NEXT:  entry:
1274 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1275 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1276 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
1277 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1278 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1279 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1280 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
1281 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1282 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1283 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1284 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1285 // CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
1286 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1287 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1288 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1289 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1290 // CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
1291 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1292 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1293 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1294 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1295 // CHECK1-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
1296 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1297 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
1298 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
1299 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
1300 // CHECK1-NEXT:    ret i32 [[TMP8]]
1301 //
1302 //
1303 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1304 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1305 // CHECK1-NEXT:  entry:
1306 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1307 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1308 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
1309 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1310 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1311 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1312 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
1313 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
1314 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
1315 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
1316 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1317 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1318 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1319 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1320 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1321 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1322 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
1323 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1324 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1325 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1326 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1327 // CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
1328 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
1329 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1330 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1331 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1332 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
1333 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1334 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
1335 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
1336 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1337 // CHECK1:       omp_if.then:
1338 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1339 // CHECK1-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
1340 // CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
1341 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1342 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i64 40, i1 false)
1343 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1344 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
1345 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
1346 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1347 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
1348 // CHECK1-NEXT:    store double* [[A]], double** [[TMP14]], align 8
1349 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1350 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
1351 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1352 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
1353 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
1354 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1355 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
1356 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
1357 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1358 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
1359 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1360 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
1361 // CHECK1-NEXT:    store i64 2, i64* [[TMP22]], align 8
1362 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1363 // CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
1364 // CHECK1-NEXT:    store i64 2, i64* [[TMP24]], align 8
1365 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1366 // CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
1367 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1368 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
1369 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
1370 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1371 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
1372 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
1373 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1374 // CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
1375 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1376 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
1377 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
1378 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1379 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
1380 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
1381 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1382 // CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
1383 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
1384 // CHECK1-NEXT:    store i8* null, i8** [[TMP36]], align 8
1385 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1386 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1387 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1388 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
1389 // CHECK1-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1390 // CHECK1-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
1391 // CHECK1-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1392 // CHECK1:       omp_offload.failed:
1393 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
1394 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1395 // CHECK1:       omp_offload.cont:
1396 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1397 // CHECK1:       omp_if.else:
1398 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
1399 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1400 // CHECK1:       omp_if.end:
1401 // CHECK1-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
1402 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
1403 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1404 // CHECK1-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
1405 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
1406 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
1407 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
1408 // CHECK1-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1409 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
1410 // CHECK1-NEXT:    ret i32 [[ADD4]]
1411 //
1412 //
1413 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
1414 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1415 // CHECK1-NEXT:  entry:
1416 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1417 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1418 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1419 // CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
1420 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1421 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1422 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1423 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1424 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1425 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
1426 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
1427 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
1428 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1429 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1430 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1431 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
1432 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1433 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1434 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1435 // CHECK1-NEXT:    store i8 0, i8* [[AAA]], align 1
1436 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1437 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1438 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1439 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[N_CASTED]], align 8
1440 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
1441 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1442 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
1443 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
1444 // CHECK1-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
1445 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1446 // CHECK1-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
1447 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1448 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
1449 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1450 // CHECK1-NEXT:    store i8 [[TMP6]], i8* [[CONV3]], align 1
1451 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1452 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
1453 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
1454 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1455 // CHECK1:       omp_if.then:
1456 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1457 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1458 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
1459 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1460 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
1461 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP12]], align 8
1462 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1463 // CHECK1-NEXT:    store i8* null, i8** [[TMP13]], align 8
1464 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1465 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1466 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
1467 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1468 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
1469 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP17]], align 8
1470 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1471 // CHECK1-NEXT:    store i8* null, i8** [[TMP18]], align 8
1472 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1473 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
1474 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
1475 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1476 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
1477 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
1478 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1479 // CHECK1-NEXT:    store i8* null, i8** [[TMP23]], align 8
1480 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1481 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
1482 // CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP25]], align 8
1483 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1484 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
1485 // CHECK1-NEXT:    store i64 [[TMP7]], i64* [[TMP27]], align 8
1486 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1487 // CHECK1-NEXT:    store i8* null, i8** [[TMP28]], align 8
1488 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1489 // CHECK1-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
1490 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 8
1491 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1492 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
1493 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 8
1494 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
1495 // CHECK1-NEXT:    store i8* null, i8** [[TMP33]], align 8
1496 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1497 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1498 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
1499 // CHECK1-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
1500 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
1501 // CHECK1-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_4]], align 4
1502 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1503 // CHECK1-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1504 // CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
1505 // CHECK1-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
1506 // CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
1507 // CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
1508 // CHECK1-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
1509 // CHECK1-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
1510 // CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
1511 // CHECK1-NEXT:    [[ADD8:%.*]] = add i32 [[TMP40]], 1
1512 // CHECK1-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD8]] to i64
1513 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
1514 // CHECK1-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1515 // CHECK1-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
1516 // CHECK1-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1517 // CHECK1:       omp_offload.failed:
1518 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
1519 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1520 // CHECK1:       omp_offload.cont:
1521 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1522 // CHECK1:       omp_if.else:
1523 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
1524 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1525 // CHECK1:       omp_if.end:
1526 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
1527 // CHECK1-NEXT:    ret i32 [[TMP44]]
1528 //
1529 //
1530 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1531 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
1532 // CHECK1-NEXT:  entry:
1533 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1534 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1535 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1536 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1537 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1538 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1539 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1540 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1541 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1542 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1543 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1544 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1545 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1546 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1547 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1548 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1549 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1550 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1551 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1552 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1553 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1554 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1555 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
1556 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1557 // CHECK1:       omp_if.then:
1558 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1559 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
1560 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
1561 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1562 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1563 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1564 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1565 // CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
1566 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1567 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
1568 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
1569 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1570 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1571 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1572 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1573 // CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
1574 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1575 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
1576 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
1577 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1578 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
1579 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
1580 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1581 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
1582 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1583 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1584 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
1585 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1586 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
1587 // CHECK1-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1588 // CHECK1:       omp_offload.failed:
1589 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1590 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1591 // CHECK1:       omp_offload.cont:
1592 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1593 // CHECK1:       omp_if.else:
1594 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1595 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1596 // CHECK1:       omp_if.end:
1597 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
1598 // CHECK1-NEXT:    ret i32 [[TMP24]]
1599 //
1600 //
1601 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
1602 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
1603 // CHECK1-NEXT:  entry:
1604 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1605 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1606 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1607 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1608 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1609 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1610 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1611 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1612 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1613 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1614 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1615 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1616 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1617 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1618 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1619 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1620 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
1621 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1622 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
1623 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
1624 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
1625 // CHECK1-NEXT:    ret void
1626 //
1627 //
1628 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
1629 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
1630 // CHECK1-NEXT:  entry:
1631 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1632 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1633 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1634 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1635 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1636 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1637 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1638 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1639 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1640 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1641 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1642 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1643 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1644 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1645 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1646 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1647 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1648 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1649 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1650 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1651 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1652 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1653 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1654 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1655 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1656 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1657 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1658 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
1659 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1660 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1661 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1662 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
1663 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1664 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1665 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
1666 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1667 // CHECK1:       cond.true:
1668 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1669 // CHECK1:       cond.false:
1670 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1671 // CHECK1-NEXT:    br label [[COND_END]]
1672 // CHECK1:       cond.end:
1673 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
1674 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1675 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1676 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
1677 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1678 // CHECK1:       omp.inner.for.cond:
1679 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1680 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1681 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
1682 // CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1683 // CHECK1:       omp.inner.for.body:
1684 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1685 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
1686 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1687 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1688 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
1689 // CHECK1-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
1690 // CHECK1-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
1691 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1692 // CHECK1-NEXT:    store double [[ADD5]], double* [[A]], align 8
1693 // CHECK1-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
1694 // CHECK1-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
1695 // CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
1696 // CHECK1-NEXT:    store double [[INC]], double* [[A6]], align 8
1697 // CHECK1-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
1698 // CHECK1-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
1699 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
1700 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1701 // CHECK1-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
1702 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1703 // CHECK1:       omp.body.continue:
1704 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1705 // CHECK1:       omp.inner.for.inc:
1706 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1707 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
1708 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
1709 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1710 // CHECK1:       omp.inner.for.end:
1711 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1712 // CHECK1:       omp.loop.exit:
1713 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
1714 // CHECK1-NEXT:    ret void
1715 //
1716 //
1717 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
1718 // CHECK1-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1719 // CHECK1-NEXT:  entry:
1720 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1721 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1722 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1723 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1724 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1725 // CHECK1-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
1726 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1727 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1728 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1729 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
1730 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1731 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1732 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1733 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1734 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1735 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1736 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1737 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1738 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1739 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1740 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1741 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
1742 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
1743 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
1744 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1745 // CHECK1-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
1746 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
1747 // CHECK1-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
1748 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1749 // CHECK1-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
1750 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1751 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
1752 // CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1753 // CHECK1-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
1754 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1755 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
1756 // CHECK1-NEXT:    ret void
1757 //
1758 //
1759 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
1760 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1761 // CHECK1-NEXT:  entry:
1762 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1763 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1764 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
1765 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1766 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1767 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1768 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1769 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1770 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1771 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1772 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1773 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
1774 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1775 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1776 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1777 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1778 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1779 // CHECK1-NEXT:    [[I8:%.*]] = alloca i32, align 4
1780 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1781 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1782 // CHECK1-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
1783 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1784 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1785 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1786 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1787 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1788 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1789 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1790 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1791 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1792 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
1793 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
1794 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
1795 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
1796 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1797 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1798 // CHECK1-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
1799 // CHECK1-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
1800 // CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
1801 // CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
1802 // CHECK1-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
1803 // CHECK1-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
1804 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1805 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
1806 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1807 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1808 // CHECK1-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
1809 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1810 // CHECK1:       omp.precond.then:
1811 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1812 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
1813 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
1814 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1815 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1816 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1817 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1818 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1819 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1820 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
1821 // CHECK1-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
1822 // CHECK1-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1823 // CHECK1:       cond.true:
1824 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
1825 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1826 // CHECK1:       cond.false:
1827 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1828 // CHECK1-NEXT:    br label [[COND_END]]
1829 // CHECK1:       cond.end:
1830 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
1831 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1832 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1833 // CHECK1-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
1834 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1835 // CHECK1:       omp.inner.for.cond:
1836 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1837 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1838 // CHECK1-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
1839 // CHECK1-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
1840 // CHECK1-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1841 // CHECK1:       omp.inner.for.body:
1842 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1843 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1844 // CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
1845 // CHECK1-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
1846 // CHECK1-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
1847 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
1848 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
1849 // CHECK1-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
1850 // CHECK1-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
1851 // CHECK1-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
1852 // CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
1853 // CHECK1-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
1854 // CHECK1-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
1855 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
1856 // CHECK1-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
1857 // CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
1858 // CHECK1-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
1859 // CHECK1-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
1860 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
1861 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1862 // CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
1863 // CHECK1-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
1864 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1865 // CHECK1:       omp.body.continue:
1866 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1867 // CHECK1:       omp.inner.for.inc:
1868 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1869 // CHECK1-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
1870 // CHECK1-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
1871 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1872 // CHECK1:       omp.inner.for.end:
1873 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1874 // CHECK1:       omp.loop.exit:
1875 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1876 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
1877 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
1878 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
1879 // CHECK1:       omp.precond.end:
1880 // CHECK1-NEXT:    ret void
1881 //
1882 //
1883 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
1884 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1885 // CHECK1-NEXT:  entry:
1886 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1887 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1888 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1889 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1890 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1891 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1892 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1893 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1894 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1895 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1896 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1897 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1898 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1899 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
1900 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1901 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
1902 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1903 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
1904 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1905 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
1906 // CHECK1-NEXT:    ret void
1907 //
1908 //
1909 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..18
1910 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1911 // CHECK1-NEXT:  entry:
1912 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1913 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1914 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1915 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1916 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1917 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1918 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1919 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1920 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1921 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1922 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1923 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1924 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1925 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1926 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1927 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1928 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1929 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1930 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1931 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1932 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1933 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
1934 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1935 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1936 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1937 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1938 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1939 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1940 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1941 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1942 // CHECK1:       cond.true:
1943 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1944 // CHECK1:       cond.false:
1945 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1946 // CHECK1-NEXT:    br label [[COND_END]]
1947 // CHECK1:       cond.end:
1948 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1949 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1950 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1951 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1952 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1953 // CHECK1:       omp.inner.for.cond:
1954 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1955 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1956 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1957 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1958 // CHECK1:       omp.inner.for.body:
1959 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1960 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1961 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1962 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1963 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
1964 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
1965 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
1966 // CHECK1-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
1967 // CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
1968 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
1969 // CHECK1-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
1970 // CHECK1-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
1971 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
1972 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1973 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
1974 // CHECK1-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
1975 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1976 // CHECK1:       omp.body.continue:
1977 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1978 // CHECK1:       omp.inner.for.inc:
1979 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1980 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
1981 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
1982 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1983 // CHECK1:       omp.inner.for.end:
1984 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1985 // CHECK1:       omp.loop.exit:
1986 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1987 // CHECK1-NEXT:    ret void
1988 //
1989 //
1990 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1991 // CHECK1-SAME: () #[[ATTR4]] {
1992 // CHECK1-NEXT:  entry:
1993 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1994 // CHECK1-NEXT:    ret void
1995 //
1996 //
1997 // CHECK2-LABEL: define {{[^@]+}}@_Z3fooi
1998 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1999 // CHECK2-NEXT:  entry:
2000 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2001 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2002 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2003 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
2004 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
2005 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2006 // CHECK2-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
2007 // CHECK2-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
2008 // CHECK2-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
2009 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2010 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2011 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2012 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2013 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
2014 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2015 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2016 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2017 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
2018 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2019 // CHECK2-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
2020 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
2021 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
2022 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
2023 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2024 // CHECK2-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
2025 // CHECK2-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
2026 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
2027 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
2028 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
2029 // CHECK2-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
2030 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_22:%.*]] = alloca i32, align 4
2031 // CHECK2-NEXT:    [[A_CASTED23:%.*]] = alloca i64, align 8
2032 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED25:%.*]] = alloca i64, align 8
2033 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [10 x i8*], align 8
2034 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [10 x i8*], align 8
2035 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [10 x i8*], align 8
2036 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
2037 // CHECK2-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
2038 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
2039 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2040 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2041 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
2042 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
2043 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
2044 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
2045 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
2046 // CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
2047 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
2048 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2049 // CHECK2-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
2050 // CHECK2-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
2051 // CHECK2-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
2052 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
2053 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2054 // CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
2055 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
2056 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2057 // CHECK2-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
2058 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2059 // CHECK2-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
2060 // CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2061 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2062 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2063 // CHECK2-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
2064 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2065 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2066 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
2067 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
2068 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
2069 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2070 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
2071 // CHECK2-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
2072 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2073 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
2074 // CHECK2-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
2075 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2076 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
2077 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2078 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
2079 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
2080 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2081 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
2082 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
2083 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2084 // CHECK2-NEXT:    store i8* null, i8** [[TMP24]], align 8
2085 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2086 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
2087 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
2088 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2089 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
2090 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
2091 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2092 // CHECK2-NEXT:    store i8* null, i8** [[TMP29]], align 8
2093 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2094 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2095 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
2096 // CHECK2-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
2097 // CHECK2-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
2098 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
2099 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2100 // CHECK2-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
2101 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
2102 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2103 // CHECK2-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
2104 // CHECK2-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 120, i64 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
2105 // CHECK2-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
2106 // CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
2107 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
2108 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
2109 // CHECK2-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
2110 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
2111 // CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
2112 // CHECK2-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
2113 // CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
2114 // CHECK2-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
2115 // CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
2116 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
2117 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
2118 // CHECK2-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
2119 // CHECK2-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
2120 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
2121 // CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
2122 // CHECK2-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
2123 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP53]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
2124 // CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
2125 // CHECK2-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
2126 // CHECK2-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
2127 // CHECK2-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP38]])
2128 // CHECK2-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
2129 // CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2130 // CHECK2-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
2131 // CHECK2-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
2132 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
2133 // CHECK2-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
2134 // CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
2135 // CHECK2-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
2136 // CHECK2-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
2137 // CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
2138 // CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
2139 // CHECK2-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
2140 // CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
2141 // CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
2142 // CHECK2-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
2143 // CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
2144 // CHECK2-NEXT:    store i8* null, i8** [[TMP65]], align 8
2145 // CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
2146 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
2147 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
2148 // CHECK2-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2149 // CHECK2-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
2150 // CHECK2-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2151 // CHECK2:       omp_offload.failed:
2152 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i64 [[TMP60]]) #[[ATTR3]]
2153 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2154 // CHECK2:       omp_offload.cont:
2155 // CHECK2-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
2156 // CHECK2-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
2157 // CHECK2-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
2158 // CHECK2-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
2159 // CHECK2-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
2160 // CHECK2-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
2161 // CHECK2-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
2162 // CHECK2-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
2163 // CHECK2-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
2164 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
2165 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2166 // CHECK2:       omp_if.then:
2167 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
2168 // CHECK2-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
2169 // CHECK2-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
2170 // CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
2171 // CHECK2-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
2172 // CHECK2-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
2173 // CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
2174 // CHECK2-NEXT:    store i8* null, i8** [[TMP79]], align 8
2175 // CHECK2-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
2176 // CHECK2-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
2177 // CHECK2-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
2178 // CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
2179 // CHECK2-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
2180 // CHECK2-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
2181 // CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
2182 // CHECK2-NEXT:    store i8* null, i8** [[TMP84]], align 8
2183 // CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
2184 // CHECK2-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
2185 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
2186 // CHECK2-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP85]], i8** [[TMP86]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2187 // CHECK2-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
2188 // CHECK2-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
2189 // CHECK2:       omp_offload.failed20:
2190 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
2191 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
2192 // CHECK2:       omp_offload.cont21:
2193 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2194 // CHECK2:       omp_if.else:
2195 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
2196 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2197 // CHECK2:       omp_if.end:
2198 // CHECK2-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
2199 // CHECK2-NEXT:    store i32 [[TMP89]], i32* [[DOTCAPTURE_EXPR_22]], align 4
2200 // CHECK2-NEXT:    [[TMP90:%.*]] = load i32, i32* [[A]], align 4
2201 // CHECK2-NEXT:    [[CONV24:%.*]] = bitcast i64* [[A_CASTED23]] to i32*
2202 // CHECK2-NEXT:    store i32 [[TMP90]], i32* [[CONV24]], align 4
2203 // CHECK2-NEXT:    [[TMP91:%.*]] = load i64, i64* [[A_CASTED23]], align 8
2204 // CHECK2-NEXT:    [[TMP92:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_22]], align 4
2205 // CHECK2-NEXT:    [[CONV26:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED25]] to i32*
2206 // CHECK2-NEXT:    store i32 [[TMP92]], i32* [[CONV26]], align 4
2207 // CHECK2-NEXT:    [[TMP93:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED25]], align 8
2208 // CHECK2-NEXT:    [[TMP94:%.*]] = load i32, i32* [[N_ADDR]], align 4
2209 // CHECK2-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[TMP94]], 20
2210 // CHECK2-NEXT:    br i1 [[CMP27]], label [[OMP_IF_THEN28:%.*]], label [[OMP_IF_ELSE35:%.*]]
2211 // CHECK2:       omp_if.then28:
2212 // CHECK2-NEXT:    [[TMP95:%.*]] = mul nuw i64 [[TMP2]], 4
2213 // CHECK2-NEXT:    [[TMP96:%.*]] = mul nuw i64 5, [[TMP5]]
2214 // CHECK2-NEXT:    [[TMP97:%.*]] = mul nuw i64 [[TMP96]], 8
2215 // CHECK2-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2216 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP98]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i64 80, i1 false)
2217 // CHECK2-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
2218 // CHECK2-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i64*
2219 // CHECK2-NEXT:    store i64 [[TMP91]], i64* [[TMP100]], align 8
2220 // CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
2221 // CHECK2-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i64*
2222 // CHECK2-NEXT:    store i64 [[TMP91]], i64* [[TMP102]], align 8
2223 // CHECK2-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
2224 // CHECK2-NEXT:    store i8* null, i8** [[TMP103]], align 8
2225 // CHECK2-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
2226 // CHECK2-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
2227 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 8
2228 // CHECK2-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
2229 // CHECK2-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
2230 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 8
2231 // CHECK2-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 1
2232 // CHECK2-NEXT:    store i8* null, i8** [[TMP108]], align 8
2233 // CHECK2-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 2
2234 // CHECK2-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i64*
2235 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP110]], align 8
2236 // CHECK2-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 2
2237 // CHECK2-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i64*
2238 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP112]], align 8
2239 // CHECK2-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 2
2240 // CHECK2-NEXT:    store i8* null, i8** [[TMP113]], align 8
2241 // CHECK2-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 3
2242 // CHECK2-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
2243 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP115]], align 8
2244 // CHECK2-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 3
2245 // CHECK2-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
2246 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP117]], align 8
2247 // CHECK2-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
2248 // CHECK2-NEXT:    store i64 [[TMP95]], i64* [[TMP118]], align 8
2249 // CHECK2-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 3
2250 // CHECK2-NEXT:    store i8* null, i8** [[TMP119]], align 8
2251 // CHECK2-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 4
2252 // CHECK2-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
2253 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 8
2254 // CHECK2-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 4
2255 // CHECK2-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
2256 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 8
2257 // CHECK2-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 4
2258 // CHECK2-NEXT:    store i8* null, i8** [[TMP124]], align 8
2259 // CHECK2-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 5
2260 // CHECK2-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i64*
2261 // CHECK2-NEXT:    store i64 5, i64* [[TMP126]], align 8
2262 // CHECK2-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 5
2263 // CHECK2-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
2264 // CHECK2-NEXT:    store i64 5, i64* [[TMP128]], align 8
2265 // CHECK2-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 5
2266 // CHECK2-NEXT:    store i8* null, i8** [[TMP129]], align 8
2267 // CHECK2-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 6
2268 // CHECK2-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
2269 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP131]], align 8
2270 // CHECK2-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 6
2271 // CHECK2-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i64*
2272 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP133]], align 8
2273 // CHECK2-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 6
2274 // CHECK2-NEXT:    store i8* null, i8** [[TMP134]], align 8
2275 // CHECK2-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 7
2276 // CHECK2-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
2277 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 8
2278 // CHECK2-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 7
2279 // CHECK2-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
2280 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 8
2281 // CHECK2-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
2282 // CHECK2-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 8
2283 // CHECK2-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 7
2284 // CHECK2-NEXT:    store i8* null, i8** [[TMP140]], align 8
2285 // CHECK2-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 8
2286 // CHECK2-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
2287 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 8
2288 // CHECK2-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 8
2289 // CHECK2-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
2290 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 8
2291 // CHECK2-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 8
2292 // CHECK2-NEXT:    store i8* null, i8** [[TMP145]], align 8
2293 // CHECK2-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 9
2294 // CHECK2-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i64*
2295 // CHECK2-NEXT:    store i64 [[TMP93]], i64* [[TMP147]], align 8
2296 // CHECK2-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 9
2297 // CHECK2-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
2298 // CHECK2-NEXT:    store i64 [[TMP93]], i64* [[TMP149]], align 8
2299 // CHECK2-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 9
2300 // CHECK2-NEXT:    store i8* null, i8** [[TMP150]], align 8
2301 // CHECK2-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
2302 // CHECK2-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
2303 // CHECK2-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2304 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
2305 // CHECK2-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2306 // CHECK2-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
2307 // CHECK2-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED33:%.*]], label [[OMP_OFFLOAD_CONT34:%.*]]
2308 // CHECK2:       omp_offload.failed33:
2309 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
2310 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT34]]
2311 // CHECK2:       omp_offload.cont34:
2312 // CHECK2-NEXT:    br label [[OMP_IF_END36:%.*]]
2313 // CHECK2:       omp_if.else35:
2314 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
2315 // CHECK2-NEXT:    br label [[OMP_IF_END36]]
2316 // CHECK2:       omp_if.end36:
2317 // CHECK2-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
2318 // CHECK2-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2319 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
2320 // CHECK2-NEXT:    ret i32 [[TMP156]]
2321 //
2322 //
2323 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
2324 // CHECK2-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
2325 // CHECK2-NEXT:  entry:
2326 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2327 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2328 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
2329 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2330 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
2331 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2332 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2333 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
2334 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2335 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2336 // CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
2337 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
2338 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
2339 // CHECK2-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
2340 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
2341 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2342 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
2343 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2344 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
2345 // CHECK2-NEXT:    ret void
2346 //
2347 //
2348 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
2349 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2350 // CHECK2-NEXT:  entry:
2351 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2352 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2353 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2354 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2355 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2356 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2357 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2358 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2359 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2360 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2361 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2362 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2363 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2364 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2365 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2366 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
2367 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2368 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2369 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2370 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2371 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2372 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2373 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
2374 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2375 // CHECK2:       cond.true:
2376 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2377 // CHECK2:       cond.false:
2378 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2379 // CHECK2-NEXT:    br label [[COND_END]]
2380 // CHECK2:       cond.end:
2381 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2382 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2383 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2384 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2385 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2386 // CHECK2:       omp.inner.for.cond:
2387 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2388 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2389 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2390 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2391 // CHECK2:       omp.inner.for.body:
2392 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2393 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
2394 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2395 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2396 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2397 // CHECK2:       omp.body.continue:
2398 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2399 // CHECK2:       omp.inner.for.inc:
2400 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2401 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
2402 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2403 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
2404 // CHECK2:       omp.inner.for.end:
2405 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2406 // CHECK2:       omp.loop.exit:
2407 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2408 // CHECK2-NEXT:    ret void
2409 //
2410 //
2411 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
2412 // CHECK2-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
2413 // CHECK2-NEXT:  entry:
2414 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
2415 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
2416 // CHECK2-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
2417 // CHECK2-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
2418 // CHECK2-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
2419 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
2420 // CHECK2-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
2421 // CHECK2-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
2422 // CHECK2-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
2423 // CHECK2-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
2424 // CHECK2-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
2425 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
2426 // CHECK2-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
2427 // CHECK2-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
2428 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
2429 // CHECK2-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
2430 // CHECK2-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
2431 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
2432 // CHECK2-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
2433 // CHECK2-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
2434 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
2435 // CHECK2-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
2436 // CHECK2-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
2437 // CHECK2-NEXT:    ret void
2438 //
2439 //
2440 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
2441 // CHECK2-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
2442 // CHECK2-NEXT:  entry:
2443 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2444 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2445 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2446 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2447 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2448 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2449 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
2450 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
2451 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
2452 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
2453 // CHECK2-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
2454 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
2455 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
2456 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2457 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2458 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2459 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2460 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2461 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2462 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2463 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2464 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2465 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2466 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2467 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
2468 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
2469 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2470 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
2471 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
2472 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
2473 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
2474 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
2475 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
2476 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
2477 // CHECK2-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
2478 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
2479 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
2480 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
2481 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
2482 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
2483 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
2484 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
2485 // CHECK2-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !21
2486 // CHECK2-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !21
2487 // CHECK2-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !21
2488 // CHECK2-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !21
2489 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
2490 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
2491 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
2492 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
2493 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
2494 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
2495 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
2496 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
2497 // CHECK2-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
2498 // CHECK2-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
2499 // CHECK2-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2500 // CHECK2:       omp_offload.failed.i:
2501 // CHECK2-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
2502 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
2503 // CHECK2-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !21
2504 // CHECK2-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !21
2505 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
2506 // CHECK2-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
2507 // CHECK2-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !21
2508 // CHECK2-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !21
2509 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
2510 // CHECK2-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
2511 // CHECK2-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !21
2512 // CHECK2-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !21
2513 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
2514 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2515 // CHECK2:       .omp_outlined..1.exit:
2516 // CHECK2-NEXT:    ret i32 0
2517 //
2518 //
2519 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
2520 // CHECK2-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
2521 // CHECK2-NEXT:  entry:
2522 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2523 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2524 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2525 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2526 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2527 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2528 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
2529 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2530 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2531 // CHECK2-NEXT:    ret void
2532 //
2533 //
2534 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
2535 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
2536 // CHECK2-NEXT:  entry:
2537 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2538 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2539 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2540 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2541 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2542 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2543 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2544 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2545 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2546 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2547 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2548 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2549 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2550 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2551 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2552 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
2553 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2554 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2555 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2556 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2557 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2558 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2559 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
2560 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2561 // CHECK2:       cond.true:
2562 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2563 // CHECK2:       cond.false:
2564 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2565 // CHECK2-NEXT:    br label [[COND_END]]
2566 // CHECK2:       cond.end:
2567 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2568 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2569 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2570 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2571 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2572 // CHECK2:       omp.inner.for.cond:
2573 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2574 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2575 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2576 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2577 // CHECK2:       omp.inner.for.body:
2578 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2579 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
2580 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2581 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2582 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
2583 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
2584 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[CONV]], align 4
2585 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2586 // CHECK2:       omp.body.continue:
2587 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2588 // CHECK2:       omp.inner.for.inc:
2589 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2590 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
2591 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2592 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
2593 // CHECK2:       omp.inner.for.end:
2594 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2595 // CHECK2:       omp.loop.exit:
2596 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2597 // CHECK2-NEXT:    ret void
2598 //
2599 //
2600 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
2601 // CHECK2-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2602 // CHECK2-NEXT:  entry:
2603 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2604 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2605 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2606 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2607 // CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
2608 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2609 // CHECK2-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
2610 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2611 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2612 // CHECK2-NEXT:    ret void
2613 //
2614 //
2615 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
2616 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2617 // CHECK2-NEXT:  entry:
2618 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2619 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2620 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2621 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2622 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2623 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2624 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2625 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2626 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2627 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2628 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2629 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2630 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2631 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2632 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2633 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
2634 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2635 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2636 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2637 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2638 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2639 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2640 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
2641 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2642 // CHECK2:       cond.true:
2643 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2644 // CHECK2:       cond.false:
2645 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2646 // CHECK2-NEXT:    br label [[COND_END]]
2647 // CHECK2:       cond.end:
2648 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2649 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2650 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2651 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2652 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2653 // CHECK2:       omp.inner.for.cond:
2654 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2655 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2656 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2657 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2658 // CHECK2:       omp.inner.for.body:
2659 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2660 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
2661 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2662 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2663 // CHECK2-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
2664 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
2665 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
2666 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
2667 // CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
2668 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2669 // CHECK2:       omp.body.continue:
2670 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2671 // CHECK2:       omp.inner.for.inc:
2672 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2673 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
2674 // CHECK2-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
2675 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
2676 // CHECK2:       omp.inner.for.end:
2677 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2678 // CHECK2:       omp.loop.exit:
2679 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2680 // CHECK2-NEXT:    ret void
2681 //
2682 //
2683 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
2684 // CHECK2-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2685 // CHECK2-NEXT:  entry:
2686 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2687 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2688 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2689 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2690 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2691 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2692 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2693 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2694 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2695 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2696 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
2697 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2698 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
2699 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2700 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
2701 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2702 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
2703 // CHECK2-NEXT:    ret void
2704 //
2705 //
2706 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
2707 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2708 // CHECK2-NEXT:  entry:
2709 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2710 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2711 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2712 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2713 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2714 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2715 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2716 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2717 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2718 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2719 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2720 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2721 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2722 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2723 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2724 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2725 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2726 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2727 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
2728 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2729 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2730 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2731 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2732 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2733 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2734 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
2735 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2736 // CHECK2:       cond.true:
2737 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2738 // CHECK2:       cond.false:
2739 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2740 // CHECK2-NEXT:    br label [[COND_END]]
2741 // CHECK2:       cond.end:
2742 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2743 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2744 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2745 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2746 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2747 // CHECK2:       omp.inner.for.cond:
2748 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2749 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2750 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2751 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2752 // CHECK2:       omp.inner.for.body:
2753 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2754 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
2755 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2756 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2757 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
2758 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
2759 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
2760 // CHECK2-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
2761 // CHECK2-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
2762 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
2763 // CHECK2-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
2764 // CHECK2-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
2765 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2766 // CHECK2:       omp.body.continue:
2767 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2768 // CHECK2:       omp.inner.for.inc:
2769 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2770 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
2771 // CHECK2-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
2772 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
2773 // CHECK2:       omp.inner.for.end:
2774 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2775 // CHECK2:       omp.loop.exit:
2776 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2777 // CHECK2-NEXT:    ret void
2778 //
2779 //
2780 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
2781 // CHECK2-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2782 // CHECK2-NEXT:  entry:
2783 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2784 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
2785 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2786 // CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
2787 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
2788 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2789 // CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
2790 // CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
2791 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
2792 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2793 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2794 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2795 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2796 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
2797 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2798 // CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
2799 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
2800 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2801 // CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
2802 // CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
2803 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
2804 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2805 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2806 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
2807 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2808 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
2809 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
2810 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2811 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
2812 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
2813 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
2814 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2815 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
2816 // CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2817 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
2818 // CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
2819 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
2820 // CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2821 // CHECK2-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
2822 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2823 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
2824 // CHECK2-NEXT:    ret void
2825 //
2826 //
2827 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
2828 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2829 // CHECK2-NEXT:  entry:
2830 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2831 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2832 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2833 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
2834 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2835 // CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
2836 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
2837 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2838 // CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
2839 // CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
2840 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
2841 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2842 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2843 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2844 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2845 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2846 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2847 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2848 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2849 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2850 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2851 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2852 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
2853 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2854 // CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
2855 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
2856 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2857 // CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
2858 // CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
2859 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
2860 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2861 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2862 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
2863 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2864 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
2865 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
2866 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2867 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
2868 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
2869 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
2870 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2871 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2872 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
2873 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2874 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2875 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
2876 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2877 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2878 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
2879 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2880 // CHECK2:       omp.dispatch.cond:
2881 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2882 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
2883 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2884 // CHECK2:       cond.true:
2885 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2886 // CHECK2:       cond.false:
2887 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2888 // CHECK2-NEXT:    br label [[COND_END]]
2889 // CHECK2:       cond.end:
2890 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2891 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2892 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2893 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2894 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2895 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2896 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2897 // CHECK2-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2898 // CHECK2:       omp.dispatch.body:
2899 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2900 // CHECK2:       omp.inner.for.cond:
2901 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
2902 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
2903 // CHECK2-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2904 // CHECK2-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2905 // CHECK2:       omp.inner.for.body:
2906 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
2907 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2908 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2909 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
2910 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !22
2911 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
2912 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !22
2913 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
2914 // CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !22
2915 // CHECK2-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
2916 // CHECK2-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
2917 // CHECK2-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
2918 // CHECK2-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !22
2919 // CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
2920 // CHECK2-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
2921 // CHECK2-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
2922 // CHECK2-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
2923 // CHECK2-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
2924 // CHECK2-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
2925 // CHECK2-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
2926 // CHECK2-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
2927 // CHECK2-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
2928 // CHECK2-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
2929 // CHECK2-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
2930 // CHECK2-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
2931 // CHECK2-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
2932 // CHECK2-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
2933 // CHECK2-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
2934 // CHECK2-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
2935 // CHECK2-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
2936 // CHECK2-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
2937 // CHECK2-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !22
2938 // CHECK2-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
2939 // CHECK2-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !22
2940 // CHECK2-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
2941 // CHECK2-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !22
2942 // CHECK2-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
2943 // CHECK2-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
2944 // CHECK2-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
2945 // CHECK2-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !22
2946 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2947 // CHECK2:       omp.body.continue:
2948 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2949 // CHECK2:       omp.inner.for.inc:
2950 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
2951 // CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
2952 // CHECK2-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
2953 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
2954 // CHECK2:       omp.inner.for.end:
2955 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2956 // CHECK2:       omp.dispatch.inc:
2957 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2958 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2959 // CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
2960 // CHECK2-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
2961 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2962 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2963 // CHECK2-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
2964 // CHECK2-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
2965 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2966 // CHECK2:       omp.dispatch.end:
2967 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
2968 // CHECK2-NEXT:    ret void
2969 //
2970 //
2971 // CHECK2-LABEL: define {{[^@]+}}@_Z3bari
2972 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
2973 // CHECK2-NEXT:  entry:
2974 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2975 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2976 // CHECK2-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
2977 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2978 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2979 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2980 // CHECK2-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
2981 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
2982 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
2983 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2984 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2985 // CHECK2-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
2986 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2987 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
2988 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
2989 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2990 // CHECK2-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
2991 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2992 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
2993 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
2994 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2995 // CHECK2-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
2996 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2997 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
2998 // CHECK2-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
2999 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
3000 // CHECK2-NEXT:    ret i32 [[TMP8]]
3001 //
3002 //
3003 // CHECK2-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
3004 // CHECK2-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
3005 // CHECK2-NEXT:  entry:
3006 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3007 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3008 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
3009 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
3010 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3011 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
3012 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
3013 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
3014 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
3015 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
3016 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3017 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3018 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3019 // CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3020 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3021 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
3022 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
3023 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3024 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
3025 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
3026 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
3027 // CHECK2-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
3028 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
3029 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
3030 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
3031 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
3032 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
3033 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
3034 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
3035 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
3036 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3037 // CHECK2:       omp_if.then:
3038 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
3039 // CHECK2-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
3040 // CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
3041 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3042 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i64 40, i1 false)
3043 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3044 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
3045 // CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
3046 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3047 // CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
3048 // CHECK2-NEXT:    store double* [[A]], double** [[TMP14]], align 8
3049 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3050 // CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
3051 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3052 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3053 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
3054 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3055 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
3056 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
3057 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3058 // CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
3059 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3060 // CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
3061 // CHECK2-NEXT:    store i64 2, i64* [[TMP22]], align 8
3062 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3063 // CHECK2-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
3064 // CHECK2-NEXT:    store i64 2, i64* [[TMP24]], align 8
3065 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3066 // CHECK2-NEXT:    store i8* null, i8** [[TMP25]], align 8
3067 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
3068 // CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
3069 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
3070 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
3071 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
3072 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
3073 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
3074 // CHECK2-NEXT:    store i8* null, i8** [[TMP30]], align 8
3075 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
3076 // CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
3077 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
3078 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
3079 // CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
3080 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
3081 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
3082 // CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
3083 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
3084 // CHECK2-NEXT:    store i8* null, i8** [[TMP36]], align 8
3085 // CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3086 // CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3087 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3088 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3089 // CHECK2-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3090 // CHECK2-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
3091 // CHECK2-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3092 // CHECK2:       omp_offload.failed:
3093 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
3094 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3095 // CHECK2:       omp_offload.cont:
3096 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
3097 // CHECK2:       omp_if.else:
3098 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
3099 // CHECK2-NEXT:    br label [[OMP_IF_END]]
3100 // CHECK2:       omp_if.end:
3101 // CHECK2-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
3102 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
3103 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
3104 // CHECK2-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
3105 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
3106 // CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
3107 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
3108 // CHECK2-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3109 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
3110 // CHECK2-NEXT:    ret i32 [[ADD4]]
3111 //
3112 //
3113 // CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
3114 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
3115 // CHECK2-NEXT:  entry:
3116 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3117 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
3118 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
3119 // CHECK2-NEXT:    [[AAA:%.*]] = alloca i8, align 1
3120 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
3121 // CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
3122 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3123 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3124 // CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
3125 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
3126 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
3127 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
3128 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3129 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3130 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
3131 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
3132 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3133 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
3134 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
3135 // CHECK2-NEXT:    store i8 0, i8* [[AAA]], align 1
3136 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3137 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3138 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
3139 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[N_CASTED]], align 8
3140 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
3141 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3142 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
3143 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
3144 // CHECK2-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
3145 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3146 // CHECK2-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
3147 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3148 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
3149 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
3150 // CHECK2-NEXT:    store i8 [[TMP6]], i8* [[CONV3]], align 1
3151 // CHECK2-NEXT:    [[TMP7:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
3152 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
3153 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
3154 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3155 // CHECK2:       omp_if.then:
3156 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3157 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3158 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
3159 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3160 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3161 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP12]], align 8
3162 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3163 // CHECK2-NEXT:    store i8* null, i8** [[TMP13]], align 8
3164 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3165 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3166 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
3167 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3168 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3169 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP17]], align 8
3170 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3171 // CHECK2-NEXT:    store i8* null, i8** [[TMP18]], align 8
3172 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3173 // CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
3174 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
3175 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3176 // CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
3177 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
3178 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3179 // CHECK2-NEXT:    store i8* null, i8** [[TMP23]], align 8
3180 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
3181 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
3182 // CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP25]], align 8
3183 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
3184 // CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
3185 // CHECK2-NEXT:    store i64 [[TMP7]], i64* [[TMP27]], align 8
3186 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
3187 // CHECK2-NEXT:    store i8* null, i8** [[TMP28]], align 8
3188 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
3189 // CHECK2-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
3190 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 8
3191 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
3192 // CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
3193 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 8
3194 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
3195 // CHECK2-NEXT:    store i8* null, i8** [[TMP33]], align 8
3196 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3197 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3198 // CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
3199 // CHECK2-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
3200 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
3201 // CHECK2-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_4]], align 4
3202 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
3203 // CHECK2-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3204 // CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
3205 // CHECK2-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
3206 // CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
3207 // CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
3208 // CHECK2-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
3209 // CHECK2-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
3210 // CHECK2-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
3211 // CHECK2-NEXT:    [[ADD8:%.*]] = add i32 [[TMP40]], 1
3212 // CHECK2-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD8]] to i64
3213 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
3214 // CHECK2-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3215 // CHECK2-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
3216 // CHECK2-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3217 // CHECK2:       omp_offload.failed:
3218 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
3219 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3220 // CHECK2:       omp_offload.cont:
3221 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
3222 // CHECK2:       omp_if.else:
3223 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
3224 // CHECK2-NEXT:    br label [[OMP_IF_END]]
3225 // CHECK2:       omp_if.end:
3226 // CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
3227 // CHECK2-NEXT:    ret i32 [[TMP44]]
3228 //
3229 //
3230 // CHECK2-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
3231 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
3232 // CHECK2-NEXT:  entry:
3233 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3234 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
3235 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
3236 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
3237 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3238 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3239 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3240 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3241 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3242 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3243 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3244 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
3245 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
3246 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
3247 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3248 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
3249 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
3250 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
3251 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3252 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
3253 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3254 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3255 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
3256 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3257 // CHECK2:       omp_if.then:
3258 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3259 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
3260 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
3261 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3262 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
3263 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
3264 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3265 // CHECK2-NEXT:    store i8* null, i8** [[TMP9]], align 8
3266 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3267 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
3268 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
3269 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3270 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
3271 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
3272 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3273 // CHECK2-NEXT:    store i8* null, i8** [[TMP14]], align 8
3274 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3275 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
3276 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
3277 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3278 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
3279 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
3280 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3281 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
3282 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3283 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3284 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3285 // CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3286 // CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
3287 // CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3288 // CHECK2:       omp_offload.failed:
3289 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
3290 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3291 // CHECK2:       omp_offload.cont:
3292 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
3293 // CHECK2:       omp_if.else:
3294 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
3295 // CHECK2-NEXT:    br label [[OMP_IF_END]]
3296 // CHECK2:       omp_if.end:
3297 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
3298 // CHECK2-NEXT:    ret i32 [[TMP24]]
3299 //
3300 //
3301 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
3302 // CHECK2-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
3303 // CHECK2-NEXT:  entry:
3304 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3305 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3306 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
3307 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
3308 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
3309 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
3310 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3311 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3312 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3313 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
3314 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
3315 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3316 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
3317 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3318 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
3319 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
3320 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
3321 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
3322 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
3323 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
3324 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
3325 // CHECK2-NEXT:    ret void
3326 //
3327 //
3328 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..12
3329 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
3330 // CHECK2-NEXT:  entry:
3331 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3332 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3333 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
3334 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
3335 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
3336 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
3337 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
3338 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3339 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3340 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3341 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3342 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3343 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3344 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
3345 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3346 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3347 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
3348 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
3349 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3350 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
3351 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
3352 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
3353 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
3354 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3355 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
3356 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
3357 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3358 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
3359 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3360 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3361 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3362 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
3363 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3364 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3365 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
3366 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3367 // CHECK2:       cond.true:
3368 // CHECK2-NEXT:    br label [[COND_END:%.*]]
3369 // CHECK2:       cond.false:
3370 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3371 // CHECK2-NEXT:    br label [[COND_END]]
3372 // CHECK2:       cond.end:
3373 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
3374 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3375 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3376 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
3377 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3378 // CHECK2:       omp.inner.for.cond:
3379 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3380 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3381 // CHECK2-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
3382 // CHECK2-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3383 // CHECK2:       omp.inner.for.body:
3384 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3385 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
3386 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3387 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3388 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
3389 // CHECK2-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
3390 // CHECK2-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
3391 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
3392 // CHECK2-NEXT:    store double [[ADD5]], double* [[A]], align 8
3393 // CHECK2-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
3394 // CHECK2-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
3395 // CHECK2-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
3396 // CHECK2-NEXT:    store double [[INC]], double* [[A6]], align 8
3397 // CHECK2-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
3398 // CHECK2-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
3399 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
3400 // CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
3401 // CHECK2-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
3402 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3403 // CHECK2:       omp.body.continue:
3404 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3405 // CHECK2:       omp.inner.for.inc:
3406 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3407 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
3408 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
3409 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
3410 // CHECK2:       omp.inner.for.end:
3411 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3412 // CHECK2:       omp.loop.exit:
3413 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
3414 // CHECK2-NEXT:    ret void
3415 //
3416 //
3417 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
3418 // CHECK2-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3419 // CHECK2-NEXT:  entry:
3420 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
3421 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3422 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3423 // CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
3424 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
3425 // CHECK2-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
3426 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3427 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3428 // CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
3429 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
3430 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3431 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3432 // CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
3433 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
3434 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3435 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3436 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3437 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
3438 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
3439 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
3440 // CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3441 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
3442 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
3443 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
3444 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3445 // CHECK2-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
3446 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
3447 // CHECK2-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
3448 // CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3449 // CHECK2-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
3450 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3451 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
3452 // CHECK2-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
3453 // CHECK2-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
3454 // CHECK2-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
3455 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
3456 // CHECK2-NEXT:    ret void
3457 //
3458 //
3459 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
3460 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3461 // CHECK2-NEXT:  entry:
3462 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3463 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3464 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
3465 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3466 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3467 // CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
3468 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
3469 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3470 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3471 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3472 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
3473 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
3474 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
3475 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3476 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3477 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3478 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3479 // CHECK2-NEXT:    [[I8:%.*]] = alloca i32, align 4
3480 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3481 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3482 // CHECK2-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
3483 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3484 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3485 // CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
3486 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
3487 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3488 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3489 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3490 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
3491 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
3492 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
3493 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
3494 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
3495 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
3496 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
3497 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3498 // CHECK2-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
3499 // CHECK2-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
3500 // CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
3501 // CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
3502 // CHECK2-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
3503 // CHECK2-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
3504 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3505 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
3506 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3507 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
3508 // CHECK2-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
3509 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3510 // CHECK2:       omp.precond.then:
3511 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3512 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
3513 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
3514 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3515 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3516 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3517 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3518 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3519 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3520 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
3521 // CHECK2-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
3522 // CHECK2-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3523 // CHECK2:       cond.true:
3524 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
3525 // CHECK2-NEXT:    br label [[COND_END:%.*]]
3526 // CHECK2:       cond.false:
3527 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3528 // CHECK2-NEXT:    br label [[COND_END]]
3529 // CHECK2:       cond.end:
3530 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3531 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3532 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3533 // CHECK2-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3534 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3535 // CHECK2:       omp.inner.for.cond:
3536 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3537 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3538 // CHECK2-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
3539 // CHECK2-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
3540 // CHECK2-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3541 // CHECK2:       omp.inner.for.body:
3542 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3543 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3544 // CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
3545 // CHECK2-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
3546 // CHECK2-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
3547 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
3548 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
3549 // CHECK2-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
3550 // CHECK2-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
3551 // CHECK2-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
3552 // CHECK2-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
3553 // CHECK2-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
3554 // CHECK2-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
3555 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
3556 // CHECK2-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
3557 // CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
3558 // CHECK2-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
3559 // CHECK2-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
3560 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
3561 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3562 // CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
3563 // CHECK2-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
3564 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3565 // CHECK2:       omp.body.continue:
3566 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3567 // CHECK2:       omp.inner.for.inc:
3568 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3569 // CHECK2-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
3570 // CHECK2-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
3571 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
3572 // CHECK2:       omp.inner.for.end:
3573 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3574 // CHECK2:       omp.loop.exit:
3575 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3576 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
3577 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
3578 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
3579 // CHECK2:       omp.precond.end:
3580 // CHECK2-NEXT:    ret void
3581 //
3582 //
3583 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
3584 // CHECK2-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3585 // CHECK2-NEXT:  entry:
3586 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3587 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3588 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
3589 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3590 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3591 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3592 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3593 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
3594 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3595 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3596 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
3597 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
3598 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3599 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
3600 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
3601 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
3602 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3603 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
3604 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3605 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
3606 // CHECK2-NEXT:    ret void
3607 //
3608 //
3609 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..18
3610 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3611 // CHECK2-NEXT:  entry:
3612 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3613 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3614 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3615 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3616 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
3617 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3618 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3619 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3620 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3621 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3622 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3623 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
3624 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3625 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3626 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3627 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3628 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
3629 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3630 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3631 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
3632 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3633 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
3634 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3635 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3636 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3637 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3638 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3639 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3640 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3641 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3642 // CHECK2:       cond.true:
3643 // CHECK2-NEXT:    br label [[COND_END:%.*]]
3644 // CHECK2:       cond.false:
3645 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3646 // CHECK2-NEXT:    br label [[COND_END]]
3647 // CHECK2:       cond.end:
3648 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3649 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3650 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3651 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3652 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3653 // CHECK2:       omp.inner.for.cond:
3654 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3655 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3656 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3657 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3658 // CHECK2:       omp.inner.for.body:
3659 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3660 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3661 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3662 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3663 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
3664 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
3665 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
3666 // CHECK2-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
3667 // CHECK2-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
3668 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
3669 // CHECK2-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
3670 // CHECK2-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
3671 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
3672 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3673 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
3674 // CHECK2-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
3675 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3676 // CHECK2:       omp.body.continue:
3677 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3678 // CHECK2:       omp.inner.for.inc:
3679 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3680 // CHECK2-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
3681 // CHECK2-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3682 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
3683 // CHECK2:       omp.inner.for.end:
3684 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3685 // CHECK2:       omp.loop.exit:
3686 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3687 // CHECK2-NEXT:    ret void
3688 //
3689 //
3690 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3691 // CHECK2-SAME: () #[[ATTR4]] {
3692 // CHECK2-NEXT:  entry:
3693 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
3694 // CHECK2-NEXT:    ret void
3695 //
3696 //
3697 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
3698 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
3699 // CHECK3-NEXT:  entry:
3700 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3701 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3702 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3703 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
3704 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
3705 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3706 // CHECK3-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
3707 // CHECK3-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
3708 // CHECK3-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
3709 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3710 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3711 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3712 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
3713 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
3714 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3715 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3716 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3717 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
3718 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3719 // CHECK3-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
3720 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
3721 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
3722 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
3723 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3724 // CHECK3-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
3725 // CHECK3-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
3726 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
3727 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
3728 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
3729 // CHECK3-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
3730 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
3731 // CHECK3-NEXT:    [[A_CASTED19:%.*]] = alloca i32, align 4
3732 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED20:%.*]] = alloca i32, align 4
3733 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [10 x i8*], align 4
3734 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [10 x i8*], align 4
3735 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [10 x i8*], align 4
3736 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
3737 // CHECK3-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
3738 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
3739 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3740 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3741 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
3742 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3743 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3744 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
3745 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
3746 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
3747 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
3748 // CHECK3-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
3749 // CHECK3-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
3750 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
3751 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3752 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
3753 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3754 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3755 // CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
3756 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3757 // CHECK3-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
3758 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3759 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3760 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3761 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3762 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3763 // CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
3764 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
3765 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3766 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
3767 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
3768 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3769 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
3770 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
3771 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3772 // CHECK3-NEXT:    store i8* null, i8** [[TMP17]], align 4
3773 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3774 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
3775 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
3776 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3777 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
3778 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
3779 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3780 // CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 4
3781 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3782 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
3783 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
3784 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3785 // CHECK3-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
3786 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
3787 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3788 // CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
3789 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3790 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3791 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
3792 // CHECK3-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
3793 // CHECK3-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
3794 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
3795 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3796 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
3797 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
3798 // CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3799 // CHECK3-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
3800 // CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 72, i32 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
3801 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
3802 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
3803 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
3804 // CHECK3-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
3805 // CHECK3-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
3806 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
3807 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
3808 // CHECK3-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
3809 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
3810 // CHECK3-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
3811 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
3812 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
3813 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
3814 // CHECK3-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
3815 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
3816 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
3817 // CHECK3-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
3818 // CHECK3-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
3819 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
3820 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
3821 // CHECK3-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
3822 // CHECK3-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
3823 // CHECK3-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP36]])
3824 // CHECK3-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
3825 // CHECK3-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
3826 // CHECK3-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
3827 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
3828 // CHECK3-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
3829 // CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
3830 // CHECK3-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
3831 // CHECK3-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
3832 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
3833 // CHECK3-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
3834 // CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
3835 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
3836 // CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
3837 // CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
3838 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
3839 // CHECK3-NEXT:    store i8* null, i8** [[TMP63]], align 4
3840 // CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
3841 // CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
3842 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3843 // CHECK3-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3844 // CHECK3-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
3845 // CHECK3-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3846 // CHECK3:       omp_offload.failed:
3847 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i32 [[TMP58]]) #[[ATTR3]]
3848 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3849 // CHECK3:       omp_offload.cont:
3850 // CHECK3-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
3851 // CHECK3-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
3852 // CHECK3-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
3853 // CHECK3-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
3854 // CHECK3-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
3855 // CHECK3-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
3856 // CHECK3-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
3857 // CHECK3-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
3858 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
3859 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3860 // CHECK3:       omp_if.then:
3861 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
3862 // CHECK3-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
3863 // CHECK3-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
3864 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
3865 // CHECK3-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
3866 // CHECK3-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
3867 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
3868 // CHECK3-NEXT:    store i8* null, i8** [[TMP77]], align 4
3869 // CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
3870 // CHECK3-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
3871 // CHECK3-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
3872 // CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
3873 // CHECK3-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
3874 // CHECK3-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
3875 // CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
3876 // CHECK3-NEXT:    store i8* null, i8** [[TMP82]], align 4
3877 // CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
3878 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
3879 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3880 // CHECK3-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP83]], i8** [[TMP84]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3881 // CHECK3-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
3882 // CHECK3-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
3883 // CHECK3:       omp_offload.failed16:
3884 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
3885 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
3886 // CHECK3:       omp_offload.cont17:
3887 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3888 // CHECK3:       omp_if.else:
3889 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
3890 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3891 // CHECK3:       omp_if.end:
3892 // CHECK3-NEXT:    [[TMP87:%.*]] = load i32, i32* [[N_ADDR]], align 4
3893 // CHECK3-NEXT:    store i32 [[TMP87]], i32* [[DOTCAPTURE_EXPR_18]], align 4
3894 // CHECK3-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A]], align 4
3895 // CHECK3-NEXT:    store i32 [[TMP88]], i32* [[A_CASTED19]], align 4
3896 // CHECK3-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A_CASTED19]], align 4
3897 // CHECK3-NEXT:    [[TMP90:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
3898 // CHECK3-NEXT:    store i32 [[TMP90]], i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
3899 // CHECK3-NEXT:    [[TMP91:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
3900 // CHECK3-NEXT:    [[TMP92:%.*]] = load i32, i32* [[N_ADDR]], align 4
3901 // CHECK3-NEXT:    [[CMP21:%.*]] = icmp sgt i32 [[TMP92]], 20
3902 // CHECK3-NEXT:    br i1 [[CMP21]], label [[OMP_IF_THEN22:%.*]], label [[OMP_IF_ELSE29:%.*]]
3903 // CHECK3:       omp_if.then22:
3904 // CHECK3-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP1]], 4
3905 // CHECK3-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
3906 // CHECK3-NEXT:    [[TMP95:%.*]] = mul nuw i32 5, [[TMP3]]
3907 // CHECK3-NEXT:    [[TMP96:%.*]] = mul nuw i32 [[TMP95]], 8
3908 // CHECK3-NEXT:    [[TMP97:%.*]] = sext i32 [[TMP96]] to i64
3909 // CHECK3-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3910 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP98]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i32 80, i1 false)
3911 // CHECK3-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
3912 // CHECK3-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
3913 // CHECK3-NEXT:    store i32 [[TMP89]], i32* [[TMP100]], align 4
3914 // CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
3915 // CHECK3-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i32*
3916 // CHECK3-NEXT:    store i32 [[TMP89]], i32* [[TMP102]], align 4
3917 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
3918 // CHECK3-NEXT:    store i8* null, i8** [[TMP103]], align 4
3919 // CHECK3-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
3920 // CHECK3-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
3921 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 4
3922 // CHECK3-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
3923 // CHECK3-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
3924 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 4
3925 // CHECK3-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
3926 // CHECK3-NEXT:    store i8* null, i8** [[TMP108]], align 4
3927 // CHECK3-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 2
3928 // CHECK3-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
3929 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP110]], align 4
3930 // CHECK3-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 2
3931 // CHECK3-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i32*
3932 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP112]], align 4
3933 // CHECK3-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 2
3934 // CHECK3-NEXT:    store i8* null, i8** [[TMP113]], align 4
3935 // CHECK3-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 3
3936 // CHECK3-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
3937 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP115]], align 4
3938 // CHECK3-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 3
3939 // CHECK3-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
3940 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP117]], align 4
3941 // CHECK3-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
3942 // CHECK3-NEXT:    store i64 [[TMP94]], i64* [[TMP118]], align 4
3943 // CHECK3-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 3
3944 // CHECK3-NEXT:    store i8* null, i8** [[TMP119]], align 4
3945 // CHECK3-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 4
3946 // CHECK3-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
3947 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 4
3948 // CHECK3-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 4
3949 // CHECK3-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
3950 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 4
3951 // CHECK3-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 4
3952 // CHECK3-NEXT:    store i8* null, i8** [[TMP124]], align 4
3953 // CHECK3-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 5
3954 // CHECK3-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i32*
3955 // CHECK3-NEXT:    store i32 5, i32* [[TMP126]], align 4
3956 // CHECK3-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 5
3957 // CHECK3-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
3958 // CHECK3-NEXT:    store i32 5, i32* [[TMP128]], align 4
3959 // CHECK3-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 5
3960 // CHECK3-NEXT:    store i8* null, i8** [[TMP129]], align 4
3961 // CHECK3-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 6
3962 // CHECK3-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i32*
3963 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP131]], align 4
3964 // CHECK3-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 6
3965 // CHECK3-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i32*
3966 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP133]], align 4
3967 // CHECK3-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 6
3968 // CHECK3-NEXT:    store i8* null, i8** [[TMP134]], align 4
3969 // CHECK3-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 7
3970 // CHECK3-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
3971 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 4
3972 // CHECK3-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 7
3973 // CHECK3-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
3974 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 4
3975 // CHECK3-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
3976 // CHECK3-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 4
3977 // CHECK3-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 7
3978 // CHECK3-NEXT:    store i8* null, i8** [[TMP140]], align 4
3979 // CHECK3-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 8
3980 // CHECK3-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
3981 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 4
3982 // CHECK3-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 8
3983 // CHECK3-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
3984 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 4
3985 // CHECK3-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 8
3986 // CHECK3-NEXT:    store i8* null, i8** [[TMP145]], align 4
3987 // CHECK3-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 9
3988 // CHECK3-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
3989 // CHECK3-NEXT:    store i32 [[TMP91]], i32* [[TMP147]], align 4
3990 // CHECK3-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 9
3991 // CHECK3-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
3992 // CHECK3-NEXT:    store i32 [[TMP91]], i32* [[TMP149]], align 4
3993 // CHECK3-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 9
3994 // CHECK3-NEXT:    store i8* null, i8** [[TMP150]], align 4
3995 // CHECK3-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
3996 // CHECK3-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
3997 // CHECK3-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3998 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3999 // CHECK3-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4000 // CHECK3-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
4001 // CHECK3-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
4002 // CHECK3:       omp_offload.failed27:
4003 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
4004 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
4005 // CHECK3:       omp_offload.cont28:
4006 // CHECK3-NEXT:    br label [[OMP_IF_END30:%.*]]
4007 // CHECK3:       omp_if.else29:
4008 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
4009 // CHECK3-NEXT:    br label [[OMP_IF_END30]]
4010 // CHECK3:       omp_if.end30:
4011 // CHECK3-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
4012 // CHECK3-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4013 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
4014 // CHECK3-NEXT:    ret i32 [[TMP156]]
4015 //
4016 //
4017 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
4018 // CHECK3-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
4019 // CHECK3-NEXT:  entry:
4020 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4021 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4022 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
4023 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4024 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
4025 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4026 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4027 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
4028 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4029 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4030 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
4031 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
4032 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
4033 // CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4034 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
4035 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4036 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
4037 // CHECK3-NEXT:    ret void
4038 //
4039 //
4040 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
4041 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4042 // CHECK3-NEXT:  entry:
4043 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4044 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4045 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4046 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4047 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4048 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4049 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4050 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4051 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4052 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4053 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4054 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4055 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4056 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4057 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4058 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
4059 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4060 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4061 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4062 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4063 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4064 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4065 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
4066 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4067 // CHECK3:       cond.true:
4068 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4069 // CHECK3:       cond.false:
4070 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4071 // CHECK3-NEXT:    br label [[COND_END]]
4072 // CHECK3:       cond.end:
4073 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4074 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4075 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4076 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4077 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4078 // CHECK3:       omp.inner.for.cond:
4079 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4080 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4081 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4082 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4083 // CHECK3:       omp.inner.for.body:
4084 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4085 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
4086 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4087 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4088 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4089 // CHECK3:       omp.body.continue:
4090 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4091 // CHECK3:       omp.inner.for.inc:
4092 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4093 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
4094 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
4095 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4096 // CHECK3:       omp.inner.for.end:
4097 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4098 // CHECK3:       omp.loop.exit:
4099 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4100 // CHECK3-NEXT:    ret void
4101 //
4102 //
4103 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_privates_map.
4104 // CHECK3-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
4105 // CHECK3-NEXT:  entry:
4106 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
4107 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
4108 // CHECK3-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
4109 // CHECK3-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
4110 // CHECK3-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
4111 // CHECK3-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
4112 // CHECK3-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
4113 // CHECK3-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
4114 // CHECK3-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
4115 // CHECK3-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
4116 // CHECK3-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
4117 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
4118 // CHECK3-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
4119 // CHECK3-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
4120 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
4121 // CHECK3-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
4122 // CHECK3-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
4123 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
4124 // CHECK3-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
4125 // CHECK3-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
4126 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
4127 // CHECK3-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
4128 // CHECK3-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
4129 // CHECK3-NEXT:    ret void
4130 //
4131 //
4132 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
4133 // CHECK3-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
4134 // CHECK3-NEXT:  entry:
4135 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4136 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
4137 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
4138 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
4139 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
4140 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
4141 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
4142 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
4143 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
4144 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
4145 // CHECK3-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
4146 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
4147 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
4148 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4149 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
4150 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4151 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
4152 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4153 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
4154 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4155 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4156 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4157 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
4158 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4159 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
4160 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
4161 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4162 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
4163 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
4164 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
4165 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
4166 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
4167 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
4168 // CHECK3-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
4169 // CHECK3-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
4170 // CHECK3-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
4171 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
4172 // CHECK3-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
4173 // CHECK3-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
4174 // CHECK3-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
4175 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
4176 // CHECK3-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
4177 // CHECK3-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !22
4178 // CHECK3-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !22
4179 // CHECK3-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !22
4180 // CHECK3-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !22
4181 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
4182 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
4183 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
4184 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
4185 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
4186 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
4187 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
4188 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
4189 // CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
4190 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4191 // CHECK3-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
4192 // CHECK3:       omp_offload.failed.i:
4193 // CHECK3-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
4194 // CHECK3-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
4195 // CHECK3-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !22
4196 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !22
4197 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
4198 // CHECK3-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
4199 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
4200 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
4201 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
4202 // CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
4203 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
4204 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
4205 // CHECK3:       .omp_outlined..1.exit:
4206 // CHECK3-NEXT:    ret i32 0
4207 //
4208 //
4209 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
4210 // CHECK3-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
4211 // CHECK3-NEXT:  entry:
4212 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4213 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4214 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4215 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4216 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4217 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4218 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
4219 // CHECK3-NEXT:    ret void
4220 //
4221 //
4222 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
4223 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
4224 // CHECK3-NEXT:  entry:
4225 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4226 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4227 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4228 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4229 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4230 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4231 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4232 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4233 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4234 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4235 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4236 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4237 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4238 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4239 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
4240 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4241 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4242 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4243 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4244 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4245 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4246 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
4247 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4248 // CHECK3:       cond.true:
4249 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4250 // CHECK3:       cond.false:
4251 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4252 // CHECK3-NEXT:    br label [[COND_END]]
4253 // CHECK3:       cond.end:
4254 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4255 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4256 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4257 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4258 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4259 // CHECK3:       omp.inner.for.cond:
4260 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4261 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4262 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4263 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4264 // CHECK3:       omp.inner.for.body:
4265 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4266 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
4267 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4268 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4269 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
4270 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
4271 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
4272 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4273 // CHECK3:       omp.body.continue:
4274 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4275 // CHECK3:       omp.inner.for.inc:
4276 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4277 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
4278 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4279 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4280 // CHECK3:       omp.inner.for.end:
4281 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4282 // CHECK3:       omp.loop.exit:
4283 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4284 // CHECK3-NEXT:    ret void
4285 //
4286 //
4287 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
4288 // CHECK3-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4289 // CHECK3-NEXT:  entry:
4290 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4291 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4292 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4293 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4294 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
4295 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4296 // CHECK3-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
4297 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4298 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
4299 // CHECK3-NEXT:    ret void
4300 //
4301 //
4302 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
4303 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4304 // CHECK3-NEXT:  entry:
4305 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4306 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4307 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4308 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4309 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4310 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4311 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4312 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4313 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4314 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4315 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4316 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4317 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4318 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4319 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4320 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
4321 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4322 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4323 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4324 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4325 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4326 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4327 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
4328 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4329 // CHECK3:       cond.true:
4330 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4331 // CHECK3:       cond.false:
4332 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4333 // CHECK3-NEXT:    br label [[COND_END]]
4334 // CHECK3:       cond.end:
4335 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4336 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4337 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4338 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4339 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4340 // CHECK3:       omp.inner.for.cond:
4341 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4342 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4343 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4344 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4345 // CHECK3:       omp.inner.for.body:
4346 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4347 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
4348 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4349 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4350 // CHECK3-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
4351 // CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
4352 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
4353 // CHECK3-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
4354 // CHECK3-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
4355 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4356 // CHECK3:       omp.body.continue:
4357 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4358 // CHECK3:       omp.inner.for.inc:
4359 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4360 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
4361 // CHECK3-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
4362 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4363 // CHECK3:       omp.inner.for.end:
4364 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4365 // CHECK3:       omp.loop.exit:
4366 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4367 // CHECK3-NEXT:    ret void
4368 //
4369 //
4370 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
4371 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4372 // CHECK3-NEXT:  entry:
4373 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4374 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4375 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4376 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4377 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4378 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4379 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4380 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4381 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4382 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4383 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
4384 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4385 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
4386 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4387 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
4388 // CHECK3-NEXT:    ret void
4389 //
4390 //
4391 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
4392 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4393 // CHECK3-NEXT:  entry:
4394 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4395 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4396 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4397 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4398 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4399 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4400 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4401 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4402 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4403 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4404 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4405 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4406 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4407 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4408 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4409 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4410 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4411 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
4412 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4413 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4414 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4415 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4416 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4417 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4418 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
4419 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4420 // CHECK3:       cond.true:
4421 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4422 // CHECK3:       cond.false:
4423 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4424 // CHECK3-NEXT:    br label [[COND_END]]
4425 // CHECK3:       cond.end:
4426 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4427 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4428 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4429 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4430 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4431 // CHECK3:       omp.inner.for.cond:
4432 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4433 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4434 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4435 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4436 // CHECK3:       omp.inner.for.body:
4437 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4438 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
4439 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4440 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4441 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
4442 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
4443 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
4444 // CHECK3-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
4445 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
4446 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
4447 // CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
4448 // CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
4449 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4450 // CHECK3:       omp.body.continue:
4451 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4452 // CHECK3:       omp.inner.for.inc:
4453 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4454 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
4455 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
4456 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
4457 // CHECK3:       omp.inner.for.end:
4458 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4459 // CHECK3:       omp.loop.exit:
4460 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4461 // CHECK3-NEXT:    ret void
4462 //
4463 //
4464 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
4465 // CHECK3-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4466 // CHECK3-NEXT:  entry:
4467 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4468 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
4469 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4470 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
4471 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
4472 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4473 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
4474 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
4475 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
4476 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4477 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4478 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4479 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4480 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
4481 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4482 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
4483 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
4484 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4485 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
4486 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
4487 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
4488 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4489 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
4490 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4491 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
4492 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
4493 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4494 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
4495 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
4496 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
4497 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
4498 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
4499 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
4500 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4501 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4502 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4503 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
4504 // CHECK3-NEXT:    ret void
4505 //
4506 //
4507 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
4508 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4509 // CHECK3-NEXT:  entry:
4510 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4511 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4512 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4513 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
4514 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4515 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
4516 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
4517 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4518 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
4519 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
4520 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
4521 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4522 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4523 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4524 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4525 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4526 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4527 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4528 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4529 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4530 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4531 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4532 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
4533 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4534 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
4535 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
4536 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4537 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
4538 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
4539 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
4540 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4541 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
4542 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4543 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
4544 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
4545 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4546 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
4547 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
4548 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
4549 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4550 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
4551 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4552 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4553 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4554 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4555 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4556 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
4557 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4558 // CHECK3:       omp.dispatch.cond:
4559 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4560 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
4561 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4562 // CHECK3:       cond.true:
4563 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4564 // CHECK3:       cond.false:
4565 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4566 // CHECK3-NEXT:    br label [[COND_END]]
4567 // CHECK3:       cond.end:
4568 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4569 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4570 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4571 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4572 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4573 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4574 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
4575 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4576 // CHECK3:       omp.dispatch.body:
4577 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4578 // CHECK3:       omp.inner.for.cond:
4579 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4580 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
4581 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
4582 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4583 // CHECK3:       omp.inner.for.body:
4584 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4585 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
4586 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4587 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
4588 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !23
4589 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
4590 // CHECK3-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !23
4591 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
4592 // CHECK3-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
4593 // CHECK3-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
4594 // CHECK3-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
4595 // CHECK3-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
4596 // CHECK3-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
4597 // CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
4598 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
4599 // CHECK3-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
4600 // CHECK3-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
4601 // CHECK3-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
4602 // CHECK3-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
4603 // CHECK3-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
4604 // CHECK3-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
4605 // CHECK3-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
4606 // CHECK3-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
4607 // CHECK3-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
4608 // CHECK3-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
4609 // CHECK3-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
4610 // CHECK3-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
4611 // CHECK3-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
4612 // CHECK3-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
4613 // CHECK3-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
4614 // CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
4615 // CHECK3-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !23
4616 // CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
4617 // CHECK3-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !23
4618 // CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
4619 // CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !23
4620 // CHECK3-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
4621 // CHECK3-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
4622 // CHECK3-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
4623 // CHECK3-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !23
4624 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4625 // CHECK3:       omp.body.continue:
4626 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4627 // CHECK3:       omp.inner.for.inc:
4628 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4629 // CHECK3-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
4630 // CHECK3-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4631 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
4632 // CHECK3:       omp.inner.for.end:
4633 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4634 // CHECK3:       omp.dispatch.inc:
4635 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4636 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4637 // CHECK3-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
4638 // CHECK3-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
4639 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4640 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4641 // CHECK3-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
4642 // CHECK3-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
4643 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
4644 // CHECK3:       omp.dispatch.end:
4645 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
4646 // CHECK3-NEXT:    ret void
4647 //
4648 //
4649 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
4650 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
4651 // CHECK3-NEXT:  entry:
4652 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4653 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
4654 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
4655 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4656 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
4657 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4658 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
4659 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
4660 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
4661 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4662 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
4663 // CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
4664 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4665 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
4666 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
4667 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4668 // CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
4669 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4670 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
4671 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
4672 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4673 // CHECK3-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
4674 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
4675 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
4676 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
4677 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
4678 // CHECK3-NEXT:    ret i32 [[TMP8]]
4679 //
4680 //
4681 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
4682 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
4683 // CHECK3-NEXT:  entry:
4684 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4685 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4686 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
4687 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
4688 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
4689 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4690 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
4691 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
4692 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
4693 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
4694 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4695 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4696 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4697 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4698 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4699 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
4700 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
4701 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
4702 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
4703 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
4704 // CHECK3-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
4705 // CHECK3-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
4706 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
4707 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
4708 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
4709 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
4710 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4711 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
4712 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4713 // CHECK3:       omp_if.then:
4714 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
4715 // CHECK3-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
4716 // CHECK3-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
4717 // CHECK3-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
4718 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
4719 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i32 40, i1 false)
4720 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4721 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
4722 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
4723 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4724 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
4725 // CHECK3-NEXT:    store double* [[A]], double** [[TMP14]], align 4
4726 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4727 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
4728 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4729 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4730 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
4731 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4732 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4733 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
4734 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4735 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
4736 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4737 // CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
4738 // CHECK3-NEXT:    store i32 2, i32* [[TMP22]], align 4
4739 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4740 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
4741 // CHECK3-NEXT:    store i32 2, i32* [[TMP24]], align 4
4742 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4743 // CHECK3-NEXT:    store i8* null, i8** [[TMP25]], align 4
4744 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4745 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
4746 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
4747 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4748 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
4749 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
4750 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4751 // CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
4752 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
4753 // CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
4754 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
4755 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
4756 // CHECK3-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
4757 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
4758 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
4759 // CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
4760 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
4761 // CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
4762 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4763 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4764 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4765 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
4766 // CHECK3-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4767 // CHECK3-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
4768 // CHECK3-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4769 // CHECK3:       omp_offload.failed:
4770 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
4771 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4772 // CHECK3:       omp_offload.cont:
4773 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4774 // CHECK3:       omp_if.else:
4775 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
4776 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4777 // CHECK3:       omp_if.end:
4778 // CHECK3-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
4779 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
4780 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
4781 // CHECK3-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
4782 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
4783 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
4784 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
4785 // CHECK3-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4786 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
4787 // CHECK3-NEXT:    ret i32 [[ADD3]]
4788 //
4789 //
4790 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
4791 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
4792 // CHECK3-NEXT:  entry:
4793 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4794 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
4795 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
4796 // CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4797 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4798 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
4799 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4800 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4801 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
4802 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
4803 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
4804 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
4805 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4806 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4807 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4808 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
4809 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4810 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
4811 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
4812 // CHECK3-NEXT:    store i8 0, i8* [[AAA]], align 1
4813 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4814 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[N_CASTED]], align 4
4815 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_CASTED]], align 4
4816 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
4817 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[A_CASTED]], align 4
4818 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
4819 // CHECK3-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
4820 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4821 // CHECK3-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
4822 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4823 // CHECK3-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
4824 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
4825 // CHECK3-NEXT:    store i8 [[TMP6]], i8* [[CONV1]], align 1
4826 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
4827 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
4828 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
4829 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4830 // CHECK3:       omp_if.then:
4831 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4832 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4833 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
4834 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4835 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4836 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP12]], align 4
4837 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4838 // CHECK3-NEXT:    store i8* null, i8** [[TMP13]], align 4
4839 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4840 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4841 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
4842 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4843 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4844 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP17]], align 4
4845 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4846 // CHECK3-NEXT:    store i8* null, i8** [[TMP18]], align 4
4847 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4848 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4849 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
4850 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4851 // CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
4852 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
4853 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4854 // CHECK3-NEXT:    store i8* null, i8** [[TMP23]], align 4
4855 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4856 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
4857 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP25]], align 4
4858 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4859 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
4860 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[TMP27]], align 4
4861 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4862 // CHECK3-NEXT:    store i8* null, i8** [[TMP28]], align 4
4863 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
4864 // CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
4865 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 4
4866 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
4867 // CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
4868 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 4
4869 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
4870 // CHECK3-NEXT:    store i8* null, i8** [[TMP33]], align 4
4871 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4872 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4873 // CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
4874 // CHECK3-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
4875 // CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
4876 // CHECK3-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4877 // CHECK3-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4878 // CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4879 // CHECK3-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
4880 // CHECK3-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
4881 // CHECK3-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
4882 // CHECK3-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
4883 // CHECK3-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
4884 // CHECK3-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
4885 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
4886 // CHECK3-NEXT:    [[ADD6:%.*]] = add i32 [[TMP40]], 1
4887 // CHECK3-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD6]] to i64
4888 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
4889 // CHECK3-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4890 // CHECK3-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
4891 // CHECK3-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4892 // CHECK3:       omp_offload.failed:
4893 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
4894 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4895 // CHECK3:       omp_offload.cont:
4896 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4897 // CHECK3:       omp_if.else:
4898 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
4899 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4900 // CHECK3:       omp_if.end:
4901 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
4902 // CHECK3-NEXT:    ret i32 [[TMP44]]
4903 //
4904 //
4905 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4906 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
4907 // CHECK3-NEXT:  entry:
4908 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4909 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
4910 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
4911 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4912 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4913 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4914 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4915 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4916 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4917 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4918 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4919 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
4920 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
4921 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4922 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4923 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4924 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4925 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4926 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
4927 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4928 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4929 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
4930 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4931 // CHECK3:       omp_if.then:
4932 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4933 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
4934 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
4935 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4936 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
4937 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
4938 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4939 // CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
4940 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4941 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
4942 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
4943 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4944 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4945 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
4946 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4947 // CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
4948 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4949 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
4950 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
4951 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4952 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
4953 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
4954 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4955 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
4956 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4957 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4958 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
4959 // CHECK3-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4960 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
4961 // CHECK3-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4962 // CHECK3:       omp_offload.failed:
4963 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4964 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4965 // CHECK3:       omp_offload.cont:
4966 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4967 // CHECK3:       omp_if.else:
4968 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4969 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4970 // CHECK3:       omp_if.end:
4971 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
4972 // CHECK3-NEXT:    ret i32 [[TMP24]]
4973 //
4974 //
4975 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
4976 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
4977 // CHECK3-NEXT:  entry:
4978 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4979 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4980 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4981 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4982 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
4983 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4984 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4985 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4986 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4987 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4988 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
4989 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4990 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4991 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4992 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
4993 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
4994 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
4995 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
4996 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
4997 // CHECK3-NEXT:    ret void
4998 //
4999 //
5000 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
5001 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
5002 // CHECK3-NEXT:  entry:
5003 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5004 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5005 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5006 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5007 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5008 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5009 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
5010 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5011 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5012 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5013 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5014 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5015 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5016 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5017 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5018 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5019 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5020 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5021 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5022 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5023 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
5024 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5025 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5026 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5027 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
5028 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5029 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
5030 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5031 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5032 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5033 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
5034 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5035 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5036 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
5037 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5038 // CHECK3:       cond.true:
5039 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5040 // CHECK3:       cond.false:
5041 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5042 // CHECK3-NEXT:    br label [[COND_END]]
5043 // CHECK3:       cond.end:
5044 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
5045 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5046 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5047 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
5048 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5049 // CHECK3:       omp.inner.for.cond:
5050 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5051 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5052 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
5053 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5054 // CHECK3:       omp.inner.for.body:
5055 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5056 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
5057 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5058 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5059 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
5060 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
5061 // CHECK3-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
5062 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
5063 // CHECK3-NEXT:    store double [[ADD4]], double* [[A]], align 4
5064 // CHECK3-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
5065 // CHECK3-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
5066 // CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
5067 // CHECK3-NEXT:    store double [[INC]], double* [[A5]], align 4
5068 // CHECK3-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
5069 // CHECK3-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
5070 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
5071 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
5072 // CHECK3-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
5073 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5074 // CHECK3:       omp.body.continue:
5075 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5076 // CHECK3:       omp.inner.for.inc:
5077 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5078 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
5079 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
5080 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5081 // CHECK3:       omp.inner.for.end:
5082 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5083 // CHECK3:       omp.loop.exit:
5084 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
5085 // CHECK3-NEXT:    ret void
5086 //
5087 //
5088 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
5089 // CHECK3-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5090 // CHECK3-NEXT:  entry:
5091 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5092 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5093 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5094 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
5095 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5096 // CHECK3-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
5097 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5098 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5099 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
5100 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5101 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5102 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5103 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
5104 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5105 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5106 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
5107 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5108 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
5109 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
5110 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
5111 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
5112 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
5113 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
5114 // CHECK3-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
5115 // CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5116 // CHECK3-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
5117 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5118 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
5119 // CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
5120 // CHECK3-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
5121 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
5122 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
5123 // CHECK3-NEXT:    ret void
5124 //
5125 //
5126 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
5127 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5128 // CHECK3-NEXT:  entry:
5129 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5130 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5131 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5132 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5133 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5134 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
5135 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5136 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5137 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5138 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5139 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5140 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
5141 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5142 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5143 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5144 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5145 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5146 // CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
5147 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5148 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5149 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5150 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5151 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5152 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
5153 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5154 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5155 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
5156 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5157 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5158 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
5159 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
5160 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5161 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5162 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5163 // CHECK3-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
5164 // CHECK3-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
5165 // CHECK3-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
5166 // CHECK3-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
5167 // CHECK3-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
5168 // CHECK3-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
5169 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5170 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
5171 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5172 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5173 // CHECK3-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
5174 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5175 // CHECK3:       omp.precond.then:
5176 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5177 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
5178 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
5179 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5180 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5181 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5182 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
5183 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5184 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5185 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
5186 // CHECK3-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
5187 // CHECK3-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5188 // CHECK3:       cond.true:
5189 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
5190 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5191 // CHECK3:       cond.false:
5192 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5193 // CHECK3-NEXT:    br label [[COND_END]]
5194 // CHECK3:       cond.end:
5195 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
5196 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5197 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5198 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
5199 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5200 // CHECK3:       omp.inner.for.cond:
5201 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5202 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5203 // CHECK3-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
5204 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
5205 // CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5206 // CHECK3:       omp.inner.for.body:
5207 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5208 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5209 // CHECK3-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
5210 // CHECK3-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
5211 // CHECK3-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
5212 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
5213 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
5214 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
5215 // CHECK3-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
5216 // CHECK3-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
5217 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
5218 // CHECK3-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
5219 // CHECK3-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
5220 // CHECK3-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
5221 // CHECK3-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
5222 // CHECK3-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
5223 // CHECK3-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
5224 // CHECK3-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
5225 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
5226 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5227 // CHECK3-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
5228 // CHECK3-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
5229 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5230 // CHECK3:       omp.body.continue:
5231 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5232 // CHECK3:       omp.inner.for.inc:
5233 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5234 // CHECK3-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
5235 // CHECK3-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
5236 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5237 // CHECK3:       omp.inner.for.end:
5238 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5239 // CHECK3:       omp.loop.exit:
5240 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5241 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
5242 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
5243 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
5244 // CHECK3:       omp.precond.end:
5245 // CHECK3-NEXT:    ret void
5246 //
5247 //
5248 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
5249 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5250 // CHECK3-NEXT:  entry:
5251 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5252 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5253 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5254 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5255 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5256 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5257 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5258 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5259 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5260 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5261 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5262 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
5263 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
5264 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
5265 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5266 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
5267 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5268 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
5269 // CHECK3-NEXT:    ret void
5270 //
5271 //
5272 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..18
5273 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5274 // CHECK3-NEXT:  entry:
5275 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5276 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5277 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5278 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5279 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5280 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5281 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5282 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5283 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5284 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5285 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5286 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
5287 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5288 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5289 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5290 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5291 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5292 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5293 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5294 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5295 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
5296 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5297 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5298 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5299 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5300 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5301 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5302 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
5303 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5304 // CHECK3:       cond.true:
5305 // CHECK3-NEXT:    br label [[COND_END:%.*]]
5306 // CHECK3:       cond.false:
5307 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5308 // CHECK3-NEXT:    br label [[COND_END]]
5309 // CHECK3:       cond.end:
5310 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5311 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5312 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5313 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5314 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5315 // CHECK3:       omp.inner.for.cond:
5316 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5317 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5318 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5319 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5320 // CHECK3:       omp.inner.for.body:
5321 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5322 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5323 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5324 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5325 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
5326 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
5327 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
5328 // CHECK3-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
5329 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
5330 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
5331 // CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
5332 // CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
5333 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
5334 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5335 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
5336 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
5337 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5338 // CHECK3:       omp.body.continue:
5339 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5340 // CHECK3:       omp.inner.for.inc:
5341 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5342 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
5343 // CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
5344 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
5345 // CHECK3:       omp.inner.for.end:
5346 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5347 // CHECK3:       omp.loop.exit:
5348 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5349 // CHECK3-NEXT:    ret void
5350 //
5351 //
5352 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5353 // CHECK3-SAME: () #[[ATTR4]] {
5354 // CHECK3-NEXT:  entry:
5355 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
5356 // CHECK3-NEXT:    ret void
5357 //
5358 //
5359 // CHECK4-LABEL: define {{[^@]+}}@_Z3fooi
5360 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
5361 // CHECK4-NEXT:  entry:
5362 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5363 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
5364 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
5365 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
5366 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
5367 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5368 // CHECK4-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
5369 // CHECK4-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
5370 // CHECK4-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
5371 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5372 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5373 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5374 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
5375 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
5376 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
5377 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
5378 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
5379 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
5380 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5381 // CHECK4-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
5382 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
5383 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
5384 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
5385 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5386 // CHECK4-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
5387 // CHECK4-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
5388 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
5389 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
5390 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
5391 // CHECK4-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
5392 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
5393 // CHECK4-NEXT:    [[A_CASTED19:%.*]] = alloca i32, align 4
5394 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED20:%.*]] = alloca i32, align 4
5395 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [10 x i8*], align 4
5396 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [10 x i8*], align 4
5397 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [10 x i8*], align 4
5398 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
5399 // CHECK4-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
5400 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
5401 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5402 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
5403 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
5404 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
5405 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
5406 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
5407 // CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
5408 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
5409 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
5410 // CHECK4-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
5411 // CHECK4-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
5412 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
5413 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5414 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
5415 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
5416 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5417 // CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
5418 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5419 // CHECK4-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
5420 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5421 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5422 // CHECK4-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5423 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5424 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5425 // CHECK4-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
5426 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
5427 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5428 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
5429 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
5430 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5431 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
5432 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
5433 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5434 // CHECK4-NEXT:    store i8* null, i8** [[TMP17]], align 4
5435 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
5436 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
5437 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
5438 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
5439 // CHECK4-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
5440 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
5441 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
5442 // CHECK4-NEXT:    store i8* null, i8** [[TMP22]], align 4
5443 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
5444 // CHECK4-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
5445 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
5446 // CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
5447 // CHECK4-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
5448 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
5449 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
5450 // CHECK4-NEXT:    store i8* null, i8** [[TMP27]], align 4
5451 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5452 // CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5453 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
5454 // CHECK4-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
5455 // CHECK4-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
5456 // CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
5457 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5458 // CHECK4-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
5459 // CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
5460 // CHECK4-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5461 // CHECK4-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
5462 // CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 72, i32 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
5463 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
5464 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
5465 // CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
5466 // CHECK4-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
5467 // CHECK4-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
5468 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
5469 // CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
5470 // CHECK4-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
5471 // CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
5472 // CHECK4-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
5473 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
5474 // CHECK4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
5475 // CHECK4-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
5476 // CHECK4-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
5477 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
5478 // CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
5479 // CHECK4-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
5480 // CHECK4-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
5481 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
5482 // CHECK4-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
5483 // CHECK4-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
5484 // CHECK4-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
5485 // CHECK4-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP36]])
5486 // CHECK4-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
5487 // CHECK4-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
5488 // CHECK4-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
5489 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
5490 // CHECK4-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
5491 // CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
5492 // CHECK4-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
5493 // CHECK4-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
5494 // CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
5495 // CHECK4-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
5496 // CHECK4-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
5497 // CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
5498 // CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
5499 // CHECK4-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
5500 // CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
5501 // CHECK4-NEXT:    store i8* null, i8** [[TMP63]], align 4
5502 // CHECK4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
5503 // CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
5504 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
5505 // CHECK4-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5506 // CHECK4-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
5507 // CHECK4-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5508 // CHECK4:       omp_offload.failed:
5509 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i32 [[TMP58]]) #[[ATTR3]]
5510 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5511 // CHECK4:       omp_offload.cont:
5512 // CHECK4-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
5513 // CHECK4-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
5514 // CHECK4-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
5515 // CHECK4-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
5516 // CHECK4-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
5517 // CHECK4-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
5518 // CHECK4-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
5519 // CHECK4-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
5520 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
5521 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5522 // CHECK4:       omp_if.then:
5523 // CHECK4-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
5524 // CHECK4-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
5525 // CHECK4-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
5526 // CHECK4-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
5527 // CHECK4-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
5528 // CHECK4-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
5529 // CHECK4-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
5530 // CHECK4-NEXT:    store i8* null, i8** [[TMP77]], align 4
5531 // CHECK4-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
5532 // CHECK4-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
5533 // CHECK4-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
5534 // CHECK4-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
5535 // CHECK4-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
5536 // CHECK4-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
5537 // CHECK4-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
5538 // CHECK4-NEXT:    store i8* null, i8** [[TMP82]], align 4
5539 // CHECK4-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
5540 // CHECK4-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
5541 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
5542 // CHECK4-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP83]], i8** [[TMP84]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5543 // CHECK4-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
5544 // CHECK4-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
5545 // CHECK4:       omp_offload.failed16:
5546 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
5547 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
5548 // CHECK4:       omp_offload.cont17:
5549 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5550 // CHECK4:       omp_if.else:
5551 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
5552 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5553 // CHECK4:       omp_if.end:
5554 // CHECK4-NEXT:    [[TMP87:%.*]] = load i32, i32* [[N_ADDR]], align 4
5555 // CHECK4-NEXT:    store i32 [[TMP87]], i32* [[DOTCAPTURE_EXPR_18]], align 4
5556 // CHECK4-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A]], align 4
5557 // CHECK4-NEXT:    store i32 [[TMP88]], i32* [[A_CASTED19]], align 4
5558 // CHECK4-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A_CASTED19]], align 4
5559 // CHECK4-NEXT:    [[TMP90:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
5560 // CHECK4-NEXT:    store i32 [[TMP90]], i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
5561 // CHECK4-NEXT:    [[TMP91:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
5562 // CHECK4-NEXT:    [[TMP92:%.*]] = load i32, i32* [[N_ADDR]], align 4
5563 // CHECK4-NEXT:    [[CMP21:%.*]] = icmp sgt i32 [[TMP92]], 20
5564 // CHECK4-NEXT:    br i1 [[CMP21]], label [[OMP_IF_THEN22:%.*]], label [[OMP_IF_ELSE29:%.*]]
5565 // CHECK4:       omp_if.then22:
5566 // CHECK4-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP1]], 4
5567 // CHECK4-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
5568 // CHECK4-NEXT:    [[TMP95:%.*]] = mul nuw i32 5, [[TMP3]]
5569 // CHECK4-NEXT:    [[TMP96:%.*]] = mul nuw i32 [[TMP95]], 8
5570 // CHECK4-NEXT:    [[TMP97:%.*]] = sext i32 [[TMP96]] to i64
5571 // CHECK4-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
5572 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP98]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i32 80, i1 false)
5573 // CHECK4-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
5574 // CHECK4-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
5575 // CHECK4-NEXT:    store i32 [[TMP89]], i32* [[TMP100]], align 4
5576 // CHECK4-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
5577 // CHECK4-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i32*
5578 // CHECK4-NEXT:    store i32 [[TMP89]], i32* [[TMP102]], align 4
5579 // CHECK4-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
5580 // CHECK4-NEXT:    store i8* null, i8** [[TMP103]], align 4
5581 // CHECK4-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
5582 // CHECK4-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
5583 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 4
5584 // CHECK4-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
5585 // CHECK4-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
5586 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 4
5587 // CHECK4-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
5588 // CHECK4-NEXT:    store i8* null, i8** [[TMP108]], align 4
5589 // CHECK4-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 2
5590 // CHECK4-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
5591 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP110]], align 4
5592 // CHECK4-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 2
5593 // CHECK4-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i32*
5594 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP112]], align 4
5595 // CHECK4-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 2
5596 // CHECK4-NEXT:    store i8* null, i8** [[TMP113]], align 4
5597 // CHECK4-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 3
5598 // CHECK4-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
5599 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP115]], align 4
5600 // CHECK4-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 3
5601 // CHECK4-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
5602 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP117]], align 4
5603 // CHECK4-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
5604 // CHECK4-NEXT:    store i64 [[TMP94]], i64* [[TMP118]], align 4
5605 // CHECK4-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 3
5606 // CHECK4-NEXT:    store i8* null, i8** [[TMP119]], align 4
5607 // CHECK4-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 4
5608 // CHECK4-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
5609 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 4
5610 // CHECK4-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 4
5611 // CHECK4-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
5612 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 4
5613 // CHECK4-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 4
5614 // CHECK4-NEXT:    store i8* null, i8** [[TMP124]], align 4
5615 // CHECK4-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 5
5616 // CHECK4-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i32*
5617 // CHECK4-NEXT:    store i32 5, i32* [[TMP126]], align 4
5618 // CHECK4-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 5
5619 // CHECK4-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
5620 // CHECK4-NEXT:    store i32 5, i32* [[TMP128]], align 4
5621 // CHECK4-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 5
5622 // CHECK4-NEXT:    store i8* null, i8** [[TMP129]], align 4
5623 // CHECK4-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 6
5624 // CHECK4-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i32*
5625 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP131]], align 4
5626 // CHECK4-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 6
5627 // CHECK4-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i32*
5628 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP133]], align 4
5629 // CHECK4-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 6
5630 // CHECK4-NEXT:    store i8* null, i8** [[TMP134]], align 4
5631 // CHECK4-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 7
5632 // CHECK4-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
5633 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 4
5634 // CHECK4-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 7
5635 // CHECK4-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
5636 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 4
5637 // CHECK4-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
5638 // CHECK4-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 4
5639 // CHECK4-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 7
5640 // CHECK4-NEXT:    store i8* null, i8** [[TMP140]], align 4
5641 // CHECK4-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 8
5642 // CHECK4-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
5643 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 4
5644 // CHECK4-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 8
5645 // CHECK4-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
5646 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 4
5647 // CHECK4-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 8
5648 // CHECK4-NEXT:    store i8* null, i8** [[TMP145]], align 4
5649 // CHECK4-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 9
5650 // CHECK4-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
5651 // CHECK4-NEXT:    store i32 [[TMP91]], i32* [[TMP147]], align 4
5652 // CHECK4-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 9
5653 // CHECK4-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
5654 // CHECK4-NEXT:    store i32 [[TMP91]], i32* [[TMP149]], align 4
5655 // CHECK4-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 9
5656 // CHECK4-NEXT:    store i8* null, i8** [[TMP150]], align 4
5657 // CHECK4-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
5658 // CHECK4-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
5659 // CHECK4-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
5660 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
5661 // CHECK4-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5662 // CHECK4-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
5663 // CHECK4-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
5664 // CHECK4:       omp_offload.failed27:
5665 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
5666 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
5667 // CHECK4:       omp_offload.cont28:
5668 // CHECK4-NEXT:    br label [[OMP_IF_END30:%.*]]
5669 // CHECK4:       omp_if.else29:
5670 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
5671 // CHECK4-NEXT:    br label [[OMP_IF_END30]]
5672 // CHECK4:       omp_if.end30:
5673 // CHECK4-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
5674 // CHECK4-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5675 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
5676 // CHECK4-NEXT:    ret i32 [[TMP156]]
5677 //
5678 //
5679 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
5680 // CHECK4-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
5681 // CHECK4-NEXT:  entry:
5682 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5683 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5684 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
5685 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5686 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
5687 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5688 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5689 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
5690 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5691 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5692 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
5693 // CHECK4-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
5694 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
5695 // CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5696 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
5697 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5698 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
5699 // CHECK4-NEXT:    ret void
5700 //
5701 //
5702 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
5703 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
5704 // CHECK4-NEXT:  entry:
5705 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5706 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5707 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5708 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5709 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5710 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5711 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5712 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5713 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5714 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5715 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5716 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5717 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5718 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5719 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5720 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
5721 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5722 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5723 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5724 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5725 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5726 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5727 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
5728 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5729 // CHECK4:       cond.true:
5730 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5731 // CHECK4:       cond.false:
5732 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5733 // CHECK4-NEXT:    br label [[COND_END]]
5734 // CHECK4:       cond.end:
5735 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5736 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5737 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5738 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5739 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5740 // CHECK4:       omp.inner.for.cond:
5741 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5742 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5743 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5744 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5745 // CHECK4:       omp.inner.for.body:
5746 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5747 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
5748 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5749 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5750 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5751 // CHECK4:       omp.body.continue:
5752 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5753 // CHECK4:       omp.inner.for.inc:
5754 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5755 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
5756 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
5757 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
5758 // CHECK4:       omp.inner.for.end:
5759 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5760 // CHECK4:       omp.loop.exit:
5761 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5762 // CHECK4-NEXT:    ret void
5763 //
5764 //
5765 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_privates_map.
5766 // CHECK4-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
5767 // CHECK4-NEXT:  entry:
5768 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
5769 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
5770 // CHECK4-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
5771 // CHECK4-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
5772 // CHECK4-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
5773 // CHECK4-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
5774 // CHECK4-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
5775 // CHECK4-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
5776 // CHECK4-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
5777 // CHECK4-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
5778 // CHECK4-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
5779 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
5780 // CHECK4-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
5781 // CHECK4-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
5782 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
5783 // CHECK4-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
5784 // CHECK4-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
5785 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
5786 // CHECK4-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
5787 // CHECK4-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
5788 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
5789 // CHECK4-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
5790 // CHECK4-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
5791 // CHECK4-NEXT:    ret void
5792 //
5793 //
5794 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
5795 // CHECK4-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
5796 // CHECK4-NEXT:  entry:
5797 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
5798 // CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
5799 // CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
5800 // CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
5801 // CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
5802 // CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
5803 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
5804 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
5805 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
5806 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
5807 // CHECK4-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
5808 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
5809 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
5810 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
5811 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
5812 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
5813 // CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
5814 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
5815 // CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
5816 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
5817 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
5818 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
5819 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
5820 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
5821 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
5822 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
5823 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
5824 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
5825 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
5826 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
5827 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
5828 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
5829 // CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
5830 // CHECK4-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
5831 // CHECK4-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
5832 // CHECK4-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
5833 // CHECK4-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
5834 // CHECK4-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
5835 // CHECK4-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
5836 // CHECK4-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
5837 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
5838 // CHECK4-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
5839 // CHECK4-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !22
5840 // CHECK4-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !22
5841 // CHECK4-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !22
5842 // CHECK4-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !22
5843 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
5844 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
5845 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
5846 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
5847 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
5848 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
5849 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
5850 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
5851 // CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
5852 // CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
5853 // CHECK4-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
5854 // CHECK4:       omp_offload.failed.i:
5855 // CHECK4-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
5856 // CHECK4-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
5857 // CHECK4-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !22
5858 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !22
5859 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
5860 // CHECK4-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
5861 // CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
5862 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
5863 // CHECK4-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
5864 // CHECK4-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
5865 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
5866 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
5867 // CHECK4:       .omp_outlined..1.exit:
5868 // CHECK4-NEXT:    ret i32 0
5869 //
5870 //
5871 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
5872 // CHECK4-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
5873 // CHECK4-NEXT:  entry:
5874 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5875 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5876 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5877 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
5878 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
5879 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
5880 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
5881 // CHECK4-NEXT:    ret void
5882 //
5883 //
5884 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
5885 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
5886 // CHECK4-NEXT:  entry:
5887 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5888 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5889 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5890 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5891 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5892 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5893 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5894 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5895 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5896 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5897 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5898 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5899 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5900 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5901 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
5902 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5903 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5904 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5905 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5906 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5907 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5908 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
5909 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5910 // CHECK4:       cond.true:
5911 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5912 // CHECK4:       cond.false:
5913 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5914 // CHECK4-NEXT:    br label [[COND_END]]
5915 // CHECK4:       cond.end:
5916 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5917 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5918 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5919 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5920 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5921 // CHECK4:       omp.inner.for.cond:
5922 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5923 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5924 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5925 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5926 // CHECK4:       omp.inner.for.body:
5927 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5928 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
5929 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5930 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5931 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
5932 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
5933 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
5934 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5935 // CHECK4:       omp.body.continue:
5936 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5937 // CHECK4:       omp.inner.for.inc:
5938 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5939 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
5940 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5941 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
5942 // CHECK4:       omp.inner.for.end:
5943 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5944 // CHECK4:       omp.loop.exit:
5945 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5946 // CHECK4-NEXT:    ret void
5947 //
5948 //
5949 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
5950 // CHECK4-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
5951 // CHECK4-NEXT:  entry:
5952 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5953 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5954 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5955 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5956 // CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
5957 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5958 // CHECK4-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
5959 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5960 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
5961 // CHECK4-NEXT:    ret void
5962 //
5963 //
5964 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
5965 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
5966 // CHECK4-NEXT:  entry:
5967 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5968 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5969 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5970 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5971 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5972 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5973 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5974 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5975 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5976 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5977 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5978 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5979 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5980 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5981 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5982 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
5983 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5984 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5985 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5986 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5987 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5988 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5989 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
5990 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5991 // CHECK4:       cond.true:
5992 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5993 // CHECK4:       cond.false:
5994 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5995 // CHECK4-NEXT:    br label [[COND_END]]
5996 // CHECK4:       cond.end:
5997 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5998 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5999 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6000 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6001 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6002 // CHECK4:       omp.inner.for.cond:
6003 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6004 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6005 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6006 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6007 // CHECK4:       omp.inner.for.body:
6008 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6009 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
6010 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6011 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6012 // CHECK4-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
6013 // CHECK4-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
6014 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
6015 // CHECK4-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
6016 // CHECK4-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
6017 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6018 // CHECK4:       omp.body.continue:
6019 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6020 // CHECK4:       omp.inner.for.inc:
6021 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6022 // CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
6023 // CHECK4-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
6024 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
6025 // CHECK4:       omp.inner.for.end:
6026 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6027 // CHECK4:       omp.loop.exit:
6028 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6029 // CHECK4-NEXT:    ret void
6030 //
6031 //
6032 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
6033 // CHECK4-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
6034 // CHECK4-NEXT:  entry:
6035 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6036 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6037 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6038 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6039 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6040 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6041 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6042 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
6043 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
6044 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
6045 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
6046 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6047 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
6048 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6049 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
6050 // CHECK4-NEXT:    ret void
6051 //
6052 //
6053 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
6054 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
6055 // CHECK4-NEXT:  entry:
6056 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6057 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6058 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6059 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6060 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6061 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6062 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6063 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6064 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6065 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6066 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6067 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6068 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6069 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6070 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6071 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6072 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6073 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
6074 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6075 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6076 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6077 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6078 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6079 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6080 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
6081 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6082 // CHECK4:       cond.true:
6083 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6084 // CHECK4:       cond.false:
6085 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6086 // CHECK4-NEXT:    br label [[COND_END]]
6087 // CHECK4:       cond.end:
6088 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6089 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6090 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6091 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6092 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6093 // CHECK4:       omp.inner.for.cond:
6094 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6095 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6096 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6097 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6098 // CHECK4:       omp.inner.for.body:
6099 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6100 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
6101 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6102 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6103 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
6104 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
6105 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
6106 // CHECK4-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
6107 // CHECK4-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
6108 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6109 // CHECK4-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
6110 // CHECK4-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
6111 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6112 // CHECK4:       omp.body.continue:
6113 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6114 // CHECK4:       omp.inner.for.inc:
6115 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6116 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
6117 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
6118 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
6119 // CHECK4:       omp.inner.for.end:
6120 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6121 // CHECK4:       omp.loop.exit:
6122 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6123 // CHECK4-NEXT:    ret void
6124 //
6125 //
6126 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
6127 // CHECK4-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
6128 // CHECK4-NEXT:  entry:
6129 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6130 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
6131 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6132 // CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
6133 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
6134 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6135 // CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
6136 // CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
6137 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
6138 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
6139 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6140 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
6141 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6142 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
6143 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6144 // CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
6145 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
6146 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6147 // CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
6148 // CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
6149 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
6150 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6151 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
6152 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6153 // CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
6154 // CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
6155 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6156 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
6157 // CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
6158 // CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
6159 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
6160 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
6161 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
6162 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6163 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
6164 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
6165 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
6166 // CHECK4-NEXT:    ret void
6167 //
6168 //
6169 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..9
6170 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
6171 // CHECK4-NEXT:  entry:
6172 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6173 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6174 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6175 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
6176 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6177 // CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
6178 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
6179 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6180 // CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
6181 // CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
6182 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
6183 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
6184 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6185 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6186 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6187 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6188 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6189 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6190 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6191 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6192 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6193 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6194 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
6195 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6196 // CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
6197 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
6198 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6199 // CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
6200 // CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
6201 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
6202 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6203 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
6204 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6205 // CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
6206 // CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
6207 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6208 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
6209 // CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
6210 // CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
6211 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6212 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
6213 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6214 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6215 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6216 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6217 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6218 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
6219 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6220 // CHECK4:       omp.dispatch.cond:
6221 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6222 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
6223 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6224 // CHECK4:       cond.true:
6225 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6226 // CHECK4:       cond.false:
6227 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6228 // CHECK4-NEXT:    br label [[COND_END]]
6229 // CHECK4:       cond.end:
6230 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6231 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6232 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6233 // CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6234 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6235 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6236 // CHECK4-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
6237 // CHECK4-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6238 // CHECK4:       omp.dispatch.body:
6239 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6240 // CHECK4:       omp.inner.for.cond:
6241 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
6242 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
6243 // CHECK4-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
6244 // CHECK4-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6245 // CHECK4:       omp.inner.for.body:
6246 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
6247 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
6248 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6249 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
6250 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !23
6251 // CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
6252 // CHECK4-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !23
6253 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
6254 // CHECK4-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
6255 // CHECK4-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
6256 // CHECK4-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
6257 // CHECK4-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
6258 // CHECK4-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
6259 // CHECK4-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
6260 // CHECK4-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
6261 // CHECK4-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
6262 // CHECK4-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
6263 // CHECK4-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
6264 // CHECK4-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
6265 // CHECK4-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
6266 // CHECK4-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
6267 // CHECK4-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
6268 // CHECK4-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
6269 // CHECK4-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
6270 // CHECK4-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
6271 // CHECK4-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
6272 // CHECK4-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
6273 // CHECK4-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
6274 // CHECK4-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
6275 // CHECK4-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
6276 // CHECK4-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
6277 // CHECK4-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !23
6278 // CHECK4-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
6279 // CHECK4-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !23
6280 // CHECK4-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
6281 // CHECK4-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !23
6282 // CHECK4-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
6283 // CHECK4-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
6284 // CHECK4-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
6285 // CHECK4-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !23
6286 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6287 // CHECK4:       omp.body.continue:
6288 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6289 // CHECK4:       omp.inner.for.inc:
6290 // CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
6291 // CHECK4-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
6292 // CHECK4-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
6293 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
6294 // CHECK4:       omp.inner.for.end:
6295 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
6296 // CHECK4:       omp.dispatch.inc:
6297 // CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6298 // CHECK4-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6299 // CHECK4-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
6300 // CHECK4-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
6301 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6302 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6303 // CHECK4-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
6304 // CHECK4-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
6305 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
6306 // CHECK4:       omp.dispatch.end:
6307 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
6308 // CHECK4-NEXT:    ret void
6309 //
6310 //
6311 // CHECK4-LABEL: define {{[^@]+}}@_Z3bari
6312 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
6313 // CHECK4-NEXT:  entry:
6314 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6315 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
6316 // CHECK4-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
6317 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6318 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
6319 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6320 // CHECK4-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
6321 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
6322 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
6323 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6324 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
6325 // CHECK4-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
6326 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
6327 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
6328 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
6329 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6330 // CHECK4-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
6331 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
6332 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
6333 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
6334 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
6335 // CHECK4-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
6336 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
6337 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
6338 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
6339 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
6340 // CHECK4-NEXT:    ret i32 [[TMP8]]
6341 //
6342 //
6343 // CHECK4-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
6344 // CHECK4-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
6345 // CHECK4-NEXT:  entry:
6346 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6347 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6348 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
6349 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
6350 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
6351 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
6352 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
6353 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
6354 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
6355 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
6356 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6357 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6358 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6359 // CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6360 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6361 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
6362 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
6363 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
6364 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
6365 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
6366 // CHECK4-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
6367 // CHECK4-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
6368 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
6369 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
6370 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
6371 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
6372 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
6373 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
6374 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6375 // CHECK4:       omp_if.then:
6376 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
6377 // CHECK4-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
6378 // CHECK4-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
6379 // CHECK4-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
6380 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
6381 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i32 40, i1 false)
6382 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6383 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
6384 // CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
6385 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6386 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
6387 // CHECK4-NEXT:    store double* [[A]], double** [[TMP14]], align 4
6388 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6389 // CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
6390 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6391 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
6392 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
6393 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6394 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
6395 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
6396 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6397 // CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
6398 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6399 // CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
6400 // CHECK4-NEXT:    store i32 2, i32* [[TMP22]], align 4
6401 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6402 // CHECK4-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
6403 // CHECK4-NEXT:    store i32 2, i32* [[TMP24]], align 4
6404 // CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6405 // CHECK4-NEXT:    store i8* null, i8** [[TMP25]], align 4
6406 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
6407 // CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
6408 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
6409 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
6410 // CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
6411 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
6412 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
6413 // CHECK4-NEXT:    store i8* null, i8** [[TMP30]], align 4
6414 // CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
6415 // CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
6416 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
6417 // CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
6418 // CHECK4-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
6419 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
6420 // CHECK4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
6421 // CHECK4-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
6422 // CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
6423 // CHECK4-NEXT:    store i8* null, i8** [[TMP36]], align 4
6424 // CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6425 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6426 // CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6427 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
6428 // CHECK4-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6429 // CHECK4-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
6430 // CHECK4-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6431 // CHECK4:       omp_offload.failed:
6432 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
6433 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6434 // CHECK4:       omp_offload.cont:
6435 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
6436 // CHECK4:       omp_if.else:
6437 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
6438 // CHECK4-NEXT:    br label [[OMP_IF_END]]
6439 // CHECK4:       omp_if.end:
6440 // CHECK4-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
6441 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
6442 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
6443 // CHECK4-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
6444 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
6445 // CHECK4-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
6446 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
6447 // CHECK4-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
6448 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
6449 // CHECK4-NEXT:    ret i32 [[ADD3]]
6450 //
6451 //
6452 // CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
6453 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
6454 // CHECK4-NEXT:  entry:
6455 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6456 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
6457 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
6458 // CHECK4-NEXT:    [[AAA:%.*]] = alloca i8, align 1
6459 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
6460 // CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
6461 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6462 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6463 // CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
6464 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
6465 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
6466 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
6467 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6468 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6469 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
6470 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
6471 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6472 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
6473 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
6474 // CHECK4-NEXT:    store i8 0, i8* [[AAA]], align 1
6475 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6476 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[N_CASTED]], align 4
6477 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_CASTED]], align 4
6478 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
6479 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[A_CASTED]], align 4
6480 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
6481 // CHECK4-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
6482 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6483 // CHECK4-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
6484 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6485 // CHECK4-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
6486 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
6487 // CHECK4-NEXT:    store i8 [[TMP6]], i8* [[CONV1]], align 1
6488 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
6489 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
6490 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
6491 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6492 // CHECK4:       omp_if.then:
6493 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6494 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
6495 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
6496 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6497 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
6498 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP12]], align 4
6499 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6500 // CHECK4-NEXT:    store i8* null, i8** [[TMP13]], align 4
6501 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6502 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
6503 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
6504 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6505 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
6506 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP17]], align 4
6507 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6508 // CHECK4-NEXT:    store i8* null, i8** [[TMP18]], align 4
6509 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6510 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
6511 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
6512 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6513 // CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
6514 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
6515 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6516 // CHECK4-NEXT:    store i8* null, i8** [[TMP23]], align 4
6517 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
6518 // CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
6519 // CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP25]], align 4
6520 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
6521 // CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
6522 // CHECK4-NEXT:    store i32 [[TMP7]], i32* [[TMP27]], align 4
6523 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
6524 // CHECK4-NEXT:    store i8* null, i8** [[TMP28]], align 4
6525 // CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
6526 // CHECK4-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
6527 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 4
6528 // CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
6529 // CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
6530 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 4
6531 // CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
6532 // CHECK4-NEXT:    store i8* null, i8** [[TMP33]], align 4
6533 // CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6534 // CHECK4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6535 // CHECK4-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
6536 // CHECK4-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
6537 // CHECK4-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
6538 // CHECK4-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_2]], align 4
6539 // CHECK4-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
6540 // CHECK4-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6541 // CHECK4-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
6542 // CHECK4-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
6543 // CHECK4-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
6544 // CHECK4-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
6545 // CHECK4-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
6546 // CHECK4-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
6547 // CHECK4-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
6548 // CHECK4-NEXT:    [[ADD6:%.*]] = add i32 [[TMP40]], 1
6549 // CHECK4-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD6]] to i64
6550 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
6551 // CHECK4-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6552 // CHECK4-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
6553 // CHECK4-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6554 // CHECK4:       omp_offload.failed:
6555 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
6556 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6557 // CHECK4:       omp_offload.cont:
6558 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
6559 // CHECK4:       omp_if.else:
6560 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
6561 // CHECK4-NEXT:    br label [[OMP_IF_END]]
6562 // CHECK4:       omp_if.end:
6563 // CHECK4-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
6564 // CHECK4-NEXT:    ret i32 [[TMP44]]
6565 //
6566 //
6567 // CHECK4-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
6568 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
6569 // CHECK4-NEXT:  entry:
6570 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6571 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
6572 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
6573 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
6574 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6575 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6576 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
6577 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
6578 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
6579 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6580 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6581 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
6582 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
6583 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
6584 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
6585 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
6586 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
6587 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6588 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
6589 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6590 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6591 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
6592 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6593 // CHECK4:       omp_if.then:
6594 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6595 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
6596 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
6597 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6598 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
6599 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
6600 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6601 // CHECK4-NEXT:    store i8* null, i8** [[TMP9]], align 4
6602 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6603 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
6604 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
6605 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6606 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
6607 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
6608 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6609 // CHECK4-NEXT:    store i8* null, i8** [[TMP14]], align 4
6610 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6611 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
6612 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
6613 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6614 // CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
6615 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
6616 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6617 // CHECK4-NEXT:    store i8* null, i8** [[TMP19]], align 4
6618 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6619 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6620 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
6621 // CHECK4-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6622 // CHECK4-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
6623 // CHECK4-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6624 // CHECK4:       omp_offload.failed:
6625 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
6626 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6627 // CHECK4:       omp_offload.cont:
6628 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
6629 // CHECK4:       omp_if.else:
6630 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
6631 // CHECK4-NEXT:    br label [[OMP_IF_END]]
6632 // CHECK4:       omp_if.end:
6633 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
6634 // CHECK4-NEXT:    ret i32 [[TMP24]]
6635 //
6636 //
6637 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
6638 // CHECK4-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
6639 // CHECK4-NEXT:  entry:
6640 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6641 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
6642 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6643 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6644 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
6645 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
6646 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6647 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
6648 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6649 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6650 // CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
6651 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6652 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6653 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6654 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
6655 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
6656 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
6657 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
6658 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
6659 // CHECK4-NEXT:    ret void
6660 //
6661 //
6662 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..12
6663 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
6664 // CHECK4-NEXT:  entry:
6665 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6666 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6667 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6668 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
6669 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6670 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6671 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
6672 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6673 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6674 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6675 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6676 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6677 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6678 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6679 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6680 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6681 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6682 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
6683 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6684 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6685 // CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
6686 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6687 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6688 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6689 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
6690 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6691 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
6692 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6693 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6694 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6695 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
6696 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6697 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6698 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
6699 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6700 // CHECK4:       cond.true:
6701 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6702 // CHECK4:       cond.false:
6703 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6704 // CHECK4-NEXT:    br label [[COND_END]]
6705 // CHECK4:       cond.end:
6706 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
6707 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6708 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6709 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
6710 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6711 // CHECK4:       omp.inner.for.cond:
6712 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6713 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6714 // CHECK4-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
6715 // CHECK4-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6716 // CHECK4:       omp.inner.for.body:
6717 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6718 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
6719 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6720 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6721 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
6722 // CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
6723 // CHECK4-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
6724 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
6725 // CHECK4-NEXT:    store double [[ADD4]], double* [[A]], align 4
6726 // CHECK4-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6727 // CHECK4-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
6728 // CHECK4-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
6729 // CHECK4-NEXT:    store double [[INC]], double* [[A5]], align 4
6730 // CHECK4-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
6731 // CHECK4-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
6732 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
6733 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
6734 // CHECK4-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
6735 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6736 // CHECK4:       omp.body.continue:
6737 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6738 // CHECK4:       omp.inner.for.inc:
6739 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6740 // CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
6741 // CHECK4-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
6742 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
6743 // CHECK4:       omp.inner.for.end:
6744 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6745 // CHECK4:       omp.loop.exit:
6746 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
6747 // CHECK4-NEXT:    ret void
6748 //
6749 //
6750 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
6751 // CHECK4-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6752 // CHECK4-NEXT:  entry:
6753 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6754 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6755 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6756 // CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
6757 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6758 // CHECK4-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
6759 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6760 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6761 // CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
6762 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6763 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6764 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6765 // CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
6766 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6767 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6768 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
6769 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6770 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
6771 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
6772 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
6773 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
6774 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
6775 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
6776 // CHECK4-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
6777 // CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6778 // CHECK4-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
6779 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6780 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
6781 // CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
6782 // CHECK4-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
6783 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
6784 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
6785 // CHECK4-NEXT:    ret void
6786 //
6787 //
6788 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..15
6789 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6790 // CHECK4-NEXT:  entry:
6791 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6792 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6793 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6794 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6795 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6796 // CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
6797 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6798 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6799 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6800 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6801 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
6802 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
6803 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6804 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6805 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6806 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6807 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6808 // CHECK4-NEXT:    [[I6:%.*]] = alloca i32, align 4
6809 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6810 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6811 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6812 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6813 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6814 // CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
6815 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6816 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6817 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
6818 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6819 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
6820 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
6821 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
6822 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
6823 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
6824 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6825 // CHECK4-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
6826 // CHECK4-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
6827 // CHECK4-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
6828 // CHECK4-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
6829 // CHECK4-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
6830 // CHECK4-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
6831 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6832 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
6833 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6834 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
6835 // CHECK4-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
6836 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6837 // CHECK4:       omp.precond.then:
6838 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6839 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
6840 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
6841 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6842 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6843 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6844 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6845 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6846 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6847 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
6848 // CHECK4-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
6849 // CHECK4-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6850 // CHECK4:       cond.true:
6851 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
6852 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6853 // CHECK4:       cond.false:
6854 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6855 // CHECK4-NEXT:    br label [[COND_END]]
6856 // CHECK4:       cond.end:
6857 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
6858 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6859 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6860 // CHECK4-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
6861 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6862 // CHECK4:       omp.inner.for.cond:
6863 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6864 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6865 // CHECK4-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
6866 // CHECK4-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
6867 // CHECK4-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6868 // CHECK4:       omp.inner.for.body:
6869 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6870 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6871 // CHECK4-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
6872 // CHECK4-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
6873 // CHECK4-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
6874 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
6875 // CHECK4-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
6876 // CHECK4-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
6877 // CHECK4-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
6878 // CHECK4-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
6879 // CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
6880 // CHECK4-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
6881 // CHECK4-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
6882 // CHECK4-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
6883 // CHECK4-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
6884 // CHECK4-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
6885 // CHECK4-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
6886 // CHECK4-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
6887 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
6888 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6889 // CHECK4-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
6890 // CHECK4-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
6891 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6892 // CHECK4:       omp.body.continue:
6893 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6894 // CHECK4:       omp.inner.for.inc:
6895 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6896 // CHECK4-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
6897 // CHECK4-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
6898 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
6899 // CHECK4:       omp.inner.for.end:
6900 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6901 // CHECK4:       omp.loop.exit:
6902 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6903 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
6904 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
6905 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
6906 // CHECK4:       omp.precond.end:
6907 // CHECK4-NEXT:    ret void
6908 //
6909 //
6910 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
6911 // CHECK4-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6912 // CHECK4-NEXT:  entry:
6913 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6914 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6915 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6916 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6917 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6918 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6919 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6920 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6921 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6922 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6923 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
6924 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
6925 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
6926 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
6927 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6928 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
6929 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6930 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
6931 // CHECK4-NEXT:    ret void
6932 //
6933 //
6934 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..18
6935 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6936 // CHECK4-NEXT:  entry:
6937 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6938 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6939 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6940 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6941 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6942 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6943 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6944 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6945 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6946 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6947 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6948 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6949 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6950 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6951 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6952 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6953 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6954 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6955 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6956 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6957 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
6958 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6959 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6960 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6961 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6962 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6963 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6964 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
6965 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6966 // CHECK4:       cond.true:
6967 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6968 // CHECK4:       cond.false:
6969 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6970 // CHECK4-NEXT:    br label [[COND_END]]
6971 // CHECK4:       cond.end:
6972 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6973 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6974 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6975 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6976 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6977 // CHECK4:       omp.inner.for.cond:
6978 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6979 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6980 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6981 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6982 // CHECK4:       omp.inner.for.body:
6983 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6984 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
6985 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6986 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6987 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
6988 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
6989 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
6990 // CHECK4-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
6991 // CHECK4-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
6992 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6993 // CHECK4-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
6994 // CHECK4-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
6995 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
6996 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6997 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
6998 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
6999 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7000 // CHECK4:       omp.body.continue:
7001 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7002 // CHECK4:       omp.inner.for.inc:
7003 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7004 // CHECK4-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
7005 // CHECK4-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
7006 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
7007 // CHECK4:       omp.inner.for.end:
7008 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7009 // CHECK4:       omp.loop.exit:
7010 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7011 // CHECK4-NEXT:    ret void
7012 //
7013 //
7014 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7015 // CHECK4-SAME: () #[[ATTR4]] {
7016 // CHECK4-NEXT:  entry:
7017 // CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
7018 // CHECK4-NEXT:    ret void
7019 //
7020 //
7021 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
7022 // CHECK9-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
7023 // CHECK9-NEXT:  entry:
7024 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7025 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
7026 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
7027 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7028 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
7029 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7030 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
7031 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
7032 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7033 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
7034 // CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
7035 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
7036 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
7037 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
7038 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7039 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7040 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
7041 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7042 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
7043 // CHECK9-NEXT:    ret void
7044 //
7045 //
7046 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
7047 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7048 // CHECK9-NEXT:  entry:
7049 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7050 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7051 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7052 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7053 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7054 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7055 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7056 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7057 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7058 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7059 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7060 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7061 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7062 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7063 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7064 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7065 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7066 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7067 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7068 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7069 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7070 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7071 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
7072 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7073 // CHECK9:       cond.true:
7074 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7075 // CHECK9:       cond.false:
7076 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7077 // CHECK9-NEXT:    br label [[COND_END]]
7078 // CHECK9:       cond.end:
7079 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7080 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7081 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7082 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7083 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7084 // CHECK9:       omp.inner.for.cond:
7085 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7086 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7087 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7088 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7089 // CHECK9:       omp.inner.for.body:
7090 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7091 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
7092 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7093 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7094 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7095 // CHECK9:       omp.body.continue:
7096 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7097 // CHECK9:       omp.inner.for.inc:
7098 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7099 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
7100 // CHECK9-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7101 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7102 // CHECK9:       omp.inner.for.end:
7103 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7104 // CHECK9:       omp.loop.exit:
7105 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7106 // CHECK9-NEXT:    ret void
7107 //
7108 //
7109 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
7110 // CHECK9-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7111 // CHECK9-NEXT:  entry:
7112 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7113 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7114 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7115 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7116 // CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
7117 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7118 // CHECK9-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
7119 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7120 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
7121 // CHECK9-NEXT:    ret void
7122 //
7123 //
7124 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
7125 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7126 // CHECK9-NEXT:  entry:
7127 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7128 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7129 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7130 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7131 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7132 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7133 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7134 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7135 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7136 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7137 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7138 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7139 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7140 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7141 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7142 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7143 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7144 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7145 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7146 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7147 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7148 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7149 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
7150 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7151 // CHECK9:       cond.true:
7152 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7153 // CHECK9:       cond.false:
7154 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7155 // CHECK9-NEXT:    br label [[COND_END]]
7156 // CHECK9:       cond.end:
7157 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7158 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7159 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7160 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7161 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7162 // CHECK9:       omp.inner.for.cond:
7163 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7164 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7165 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7166 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7167 // CHECK9:       omp.inner.for.body:
7168 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7169 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
7170 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7171 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7172 // CHECK9-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
7173 // CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
7174 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
7175 // CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
7176 // CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
7177 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7178 // CHECK9:       omp.body.continue:
7179 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7180 // CHECK9:       omp.inner.for.inc:
7181 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7182 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
7183 // CHECK9-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
7184 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7185 // CHECK9:       omp.inner.for.end:
7186 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7187 // CHECK9:       omp.loop.exit:
7188 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7189 // CHECK9-NEXT:    ret void
7190 //
7191 //
7192 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
7193 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7194 // CHECK9-NEXT:  entry:
7195 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7196 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7197 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7198 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7199 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7200 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7201 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7202 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7203 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
7204 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7205 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
7206 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
7207 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
7208 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7209 // CHECK9-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
7210 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7211 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
7212 // CHECK9-NEXT:    ret void
7213 //
7214 //
7215 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
7216 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7217 // CHECK9-NEXT:  entry:
7218 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7219 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7220 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7221 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7222 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7223 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7224 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7225 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7226 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7227 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7228 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7229 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7230 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7231 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7232 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7233 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7234 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7235 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7236 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7237 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7238 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7239 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7240 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7241 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7242 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7243 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
7244 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7245 // CHECK9:       cond.true:
7246 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7247 // CHECK9:       cond.false:
7248 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7249 // CHECK9-NEXT:    br label [[COND_END]]
7250 // CHECK9:       cond.end:
7251 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7252 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7253 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7254 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7255 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7256 // CHECK9:       omp.inner.for.cond:
7257 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7258 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7259 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7260 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7261 // CHECK9:       omp.inner.for.body:
7262 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7263 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
7264 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7265 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7266 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
7267 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
7268 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
7269 // CHECK9-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
7270 // CHECK9-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
7271 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
7272 // CHECK9-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
7273 // CHECK9-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
7274 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7275 // CHECK9:       omp.body.continue:
7276 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7277 // CHECK9:       omp.inner.for.inc:
7278 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7279 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
7280 // CHECK9-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
7281 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7282 // CHECK9:       omp.inner.for.end:
7283 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7284 // CHECK9:       omp.loop.exit:
7285 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7286 // CHECK9-NEXT:    ret void
7287 //
7288 //
7289 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
7290 // CHECK9-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
7291 // CHECK9-NEXT:  entry:
7292 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7293 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
7294 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7295 // CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
7296 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
7297 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7298 // CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
7299 // CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
7300 // CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
7301 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
7302 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7303 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
7304 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7305 // CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
7306 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7307 // CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
7308 // CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
7309 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7310 // CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
7311 // CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
7312 // CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
7313 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
7314 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7315 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
7316 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7317 // CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
7318 // CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
7319 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7320 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
7321 // CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
7322 // CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
7323 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
7324 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
7325 // CHECK9-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7326 // CHECK9-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
7327 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
7328 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
7329 // CHECK9-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
7330 // CHECK9-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
7331 // CHECK9-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
7332 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
7333 // CHECK9-NEXT:    ret void
7334 //
7335 //
7336 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
7337 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
7338 // CHECK9-NEXT:  entry:
7339 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7340 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7341 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7342 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
7343 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7344 // CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
7345 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
7346 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7347 // CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
7348 // CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
7349 // CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
7350 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
7351 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7352 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7353 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7354 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7355 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7356 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7357 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7358 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7359 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7360 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7361 // CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
7362 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7363 // CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
7364 // CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
7365 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7366 // CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
7367 // CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
7368 // CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
7369 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
7370 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7371 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
7372 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7373 // CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
7374 // CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
7375 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7376 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
7377 // CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
7378 // CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
7379 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
7380 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7381 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7382 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7383 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7384 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
7385 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7386 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
7387 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
7388 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
7389 // CHECK9:       omp.dispatch.cond:
7390 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7391 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
7392 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7393 // CHECK9:       cond.true:
7394 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7395 // CHECK9:       cond.false:
7396 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7397 // CHECK9-NEXT:    br label [[COND_END]]
7398 // CHECK9:       cond.end:
7399 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
7400 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7401 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7402 // CHECK9-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
7403 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7404 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7405 // CHECK9-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
7406 // CHECK9-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
7407 // CHECK9:       omp.dispatch.body:
7408 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7409 // CHECK9:       omp.inner.for.cond:
7410 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7411 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
7412 // CHECK9-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
7413 // CHECK9-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7414 // CHECK9:       omp.inner.for.body:
7415 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7416 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
7417 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7418 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
7419 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !13
7420 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
7421 // CHECK9-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !13
7422 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
7423 // CHECK9-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !13
7424 // CHECK9-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
7425 // CHECK9-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
7426 // CHECK9-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
7427 // CHECK9-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !13
7428 // CHECK9-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
7429 // CHECK9-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
7430 // CHECK9-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
7431 // CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
7432 // CHECK9-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
7433 // CHECK9-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
7434 // CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
7435 // CHECK9-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
7436 // CHECK9-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
7437 // CHECK9-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
7438 // CHECK9-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
7439 // CHECK9-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
7440 // CHECK9-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
7441 // CHECK9-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
7442 // CHECK9-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
7443 // CHECK9-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
7444 // CHECK9-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
7445 // CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
7446 // CHECK9-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !13
7447 // CHECK9-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
7448 // CHECK9-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !13
7449 // CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
7450 // CHECK9-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !13
7451 // CHECK9-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
7452 // CHECK9-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
7453 // CHECK9-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
7454 // CHECK9-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !13
7455 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7456 // CHECK9:       omp.body.continue:
7457 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7458 // CHECK9:       omp.inner.for.inc:
7459 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7460 // CHECK9-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
7461 // CHECK9-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7462 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
7463 // CHECK9:       omp.inner.for.end:
7464 // CHECK9-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
7465 // CHECK9:       omp.dispatch.inc:
7466 // CHECK9-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7467 // CHECK9-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7468 // CHECK9-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
7469 // CHECK9-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
7470 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7471 // CHECK9-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7472 // CHECK9-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
7473 // CHECK9-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
7474 // CHECK9-NEXT:    br label [[OMP_DISPATCH_COND]]
7475 // CHECK9:       omp.dispatch.end:
7476 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
7477 // CHECK9-NEXT:    ret void
7478 //
7479 //
7480 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
7481 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7482 // CHECK9-NEXT:  entry:
7483 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
7484 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7485 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7486 // CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
7487 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7488 // CHECK9-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
7489 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7490 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7491 // CHECK9-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
7492 // CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
7493 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7494 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7495 // CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
7496 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7497 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
7498 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7499 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7500 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
7501 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7502 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
7503 // CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
7504 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
7505 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
7506 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
7507 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7508 // CHECK9-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
7509 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
7510 // CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
7511 // CHECK9-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7512 // CHECK9-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
7513 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7514 // CHECK9-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
7515 // CHECK9-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
7516 // CHECK9-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
7517 // CHECK9-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
7518 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
7519 // CHECK9-NEXT:    ret void
7520 //
7521 //
7522 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
7523 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7524 // CHECK9-NEXT:  entry:
7525 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7526 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7527 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
7528 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7529 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7530 // CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
7531 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7532 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7533 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7534 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7535 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
7536 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
7537 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7538 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7539 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7540 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7541 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7542 // CHECK9-NEXT:    [[I8:%.*]] = alloca i32, align 4
7543 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7544 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7545 // CHECK9-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
7546 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7547 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7548 // CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
7549 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7550 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
7551 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7552 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7553 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
7554 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7555 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
7556 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
7557 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
7558 // CHECK9-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
7559 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
7560 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7561 // CHECK9-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
7562 // CHECK9-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
7563 // CHECK9-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
7564 // CHECK9-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
7565 // CHECK9-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
7566 // CHECK9-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
7567 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7568 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
7569 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7570 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
7571 // CHECK9-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
7572 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7573 // CHECK9:       omp.precond.then:
7574 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7575 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
7576 // CHECK9-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
7577 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7578 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7579 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7580 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
7581 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7582 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7583 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
7584 // CHECK9-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
7585 // CHECK9-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7586 // CHECK9:       cond.true:
7587 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
7588 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7589 // CHECK9:       cond.false:
7590 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7591 // CHECK9-NEXT:    br label [[COND_END]]
7592 // CHECK9:       cond.end:
7593 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
7594 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7595 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7596 // CHECK9-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
7597 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7598 // CHECK9:       omp.inner.for.cond:
7599 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7600 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7601 // CHECK9-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
7602 // CHECK9-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
7603 // CHECK9-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7604 // CHECK9:       omp.inner.for.body:
7605 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7606 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7607 // CHECK9-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
7608 // CHECK9-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
7609 // CHECK9-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
7610 // CHECK9-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
7611 // CHECK9-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
7612 // CHECK9-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
7613 // CHECK9-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
7614 // CHECK9-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
7615 // CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
7616 // CHECK9-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
7617 // CHECK9-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
7618 // CHECK9-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
7619 // CHECK9-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
7620 // CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
7621 // CHECK9-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
7622 // CHECK9-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
7623 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
7624 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7625 // CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
7626 // CHECK9-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
7627 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7628 // CHECK9:       omp.body.continue:
7629 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7630 // CHECK9:       omp.inner.for.inc:
7631 // CHECK9-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7632 // CHECK9-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
7633 // CHECK9-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
7634 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7635 // CHECK9:       omp.inner.for.end:
7636 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7637 // CHECK9:       omp.loop.exit:
7638 // CHECK9-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7639 // CHECK9-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
7640 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
7641 // CHECK9-NEXT:    br label [[OMP_PRECOND_END]]
7642 // CHECK9:       omp.precond.end:
7643 // CHECK9-NEXT:    ret void
7644 //
7645 //
7646 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
7647 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
7648 // CHECK9-NEXT:  entry:
7649 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
7650 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
7651 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7652 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7653 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
7654 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
7655 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7656 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
7657 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7658 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7659 // CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
7660 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7661 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
7662 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7663 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7664 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
7665 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
7666 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
7667 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
7668 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
7669 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
7670 // CHECK9-NEXT:    ret void
7671 //
7672 //
7673 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
7674 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
7675 // CHECK9-NEXT:  entry:
7676 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7677 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7678 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
7679 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
7680 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
7681 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
7682 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
7683 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7684 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7685 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7686 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7687 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7688 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7689 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7690 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7691 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7692 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7693 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
7694 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7695 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
7696 // CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
7697 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7698 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
7699 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7700 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
7701 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
7702 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7703 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7704 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7705 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7706 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7707 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
7708 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7709 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7710 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
7711 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7712 // CHECK9:       cond.true:
7713 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7714 // CHECK9:       cond.false:
7715 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7716 // CHECK9-NEXT:    br label [[COND_END]]
7717 // CHECK9:       cond.end:
7718 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
7719 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7720 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7721 // CHECK9-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
7722 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7723 // CHECK9:       omp.inner.for.cond:
7724 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7725 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7726 // CHECK9-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
7727 // CHECK9-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7728 // CHECK9:       omp.inner.for.body:
7729 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7730 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
7731 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7732 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7733 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
7734 // CHECK9-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
7735 // CHECK9-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
7736 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
7737 // CHECK9-NEXT:    store double [[ADD5]], double* [[A]], align 8
7738 // CHECK9-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
7739 // CHECK9-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
7740 // CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
7741 // CHECK9-NEXT:    store double [[INC]], double* [[A6]], align 8
7742 // CHECK9-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
7743 // CHECK9-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
7744 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
7745 // CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
7746 // CHECK9-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
7747 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7748 // CHECK9:       omp.body.continue:
7749 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7750 // CHECK9:       omp.inner.for.inc:
7751 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7752 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
7753 // CHECK9-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
7754 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7755 // CHECK9:       omp.inner.for.end:
7756 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7757 // CHECK9:       omp.loop.exit:
7758 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
7759 // CHECK9-NEXT:    ret void
7760 //
7761 //
7762 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
7763 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7764 // CHECK9-NEXT:  entry:
7765 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7766 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7767 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7768 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7769 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7770 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7771 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7772 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7773 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7774 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7775 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7776 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
7777 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7778 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
7779 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
7780 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
7781 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7782 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
7783 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7784 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
7785 // CHECK9-NEXT:    ret void
7786 //
7787 //
7788 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
7789 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7790 // CHECK9-NEXT:  entry:
7791 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7792 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7793 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
7794 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7795 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
7796 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7797 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7798 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7799 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7800 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7801 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7802 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7803 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7804 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7805 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
7806 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7807 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
7808 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
7809 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7810 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
7811 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7812 // CHECK9-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7813 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7814 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7815 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7816 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7817 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7818 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7819 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
7820 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7821 // CHECK9:       cond.true:
7822 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7823 // CHECK9:       cond.false:
7824 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7825 // CHECK9-NEXT:    br label [[COND_END]]
7826 // CHECK9:       cond.end:
7827 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7828 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7829 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7830 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7831 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7832 // CHECK9:       omp.inner.for.cond:
7833 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7834 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7835 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7836 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7837 // CHECK9:       omp.inner.for.body:
7838 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7839 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
7840 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7841 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7842 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
7843 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
7844 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
7845 // CHECK9-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
7846 // CHECK9-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
7847 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
7848 // CHECK9-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
7849 // CHECK9-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
7850 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
7851 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7852 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
7853 // CHECK9-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
7854 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7855 // CHECK9:       omp.body.continue:
7856 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7857 // CHECK9:       omp.inner.for.inc:
7858 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7859 // CHECK9-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
7860 // CHECK9-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
7861 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
7862 // CHECK9:       omp.inner.for.end:
7863 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7864 // CHECK9:       omp.loop.exit:
7865 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7866 // CHECK9-NEXT:    ret void
7867 //
7868 //
7869 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
7870 // CHECK10-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
7871 // CHECK10-NEXT:  entry:
7872 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7873 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
7874 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
7875 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7876 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
7877 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7878 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
7879 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
7880 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7881 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
7882 // CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
7883 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
7884 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
7885 // CHECK10-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
7886 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7887 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7888 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
7889 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7890 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
7891 // CHECK10-NEXT:    ret void
7892 //
7893 //
7894 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
7895 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7896 // CHECK10-NEXT:  entry:
7897 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7898 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7899 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7900 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7901 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7902 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7903 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7904 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7905 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7906 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
7907 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7908 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7909 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7910 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7911 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7912 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7913 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7914 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7915 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7916 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7917 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7918 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7919 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
7920 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7921 // CHECK10:       cond.true:
7922 // CHECK10-NEXT:    br label [[COND_END:%.*]]
7923 // CHECK10:       cond.false:
7924 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7925 // CHECK10-NEXT:    br label [[COND_END]]
7926 // CHECK10:       cond.end:
7927 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7928 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7929 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7930 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7931 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7932 // CHECK10:       omp.inner.for.cond:
7933 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7934 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7935 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7936 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7937 // CHECK10:       omp.inner.for.body:
7938 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7939 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
7940 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7941 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7942 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7943 // CHECK10:       omp.body.continue:
7944 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7945 // CHECK10:       omp.inner.for.inc:
7946 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7947 // CHECK10-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
7948 // CHECK10-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7949 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
7950 // CHECK10:       omp.inner.for.end:
7951 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7952 // CHECK10:       omp.loop.exit:
7953 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7954 // CHECK10-NEXT:    ret void
7955 //
7956 //
7957 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
7958 // CHECK10-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7959 // CHECK10-NEXT:  entry:
7960 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7961 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7962 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7963 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7964 // CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
7965 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7966 // CHECK10-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
7967 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7968 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
7969 // CHECK10-NEXT:    ret void
7970 //
7971 //
7972 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
7973 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
7974 // CHECK10-NEXT:  entry:
7975 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7976 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7977 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
7978 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7979 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7980 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7981 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7982 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7983 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7984 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
7985 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7986 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7987 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
7988 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
7989 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7990 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
7991 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7992 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7993 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7994 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7995 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7996 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7997 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
7998 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7999 // CHECK10:       cond.true:
8000 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8001 // CHECK10:       cond.false:
8002 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8003 // CHECK10-NEXT:    br label [[COND_END]]
8004 // CHECK10:       cond.end:
8005 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8006 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8007 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8008 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8009 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8010 // CHECK10:       omp.inner.for.cond:
8011 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8012 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8013 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8014 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8015 // CHECK10:       omp.inner.for.body:
8016 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8017 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
8018 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8019 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8020 // CHECK10-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
8021 // CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
8022 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
8023 // CHECK10-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
8024 // CHECK10-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
8025 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8026 // CHECK10:       omp.body.continue:
8027 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8028 // CHECK10:       omp.inner.for.inc:
8029 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8030 // CHECK10-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
8031 // CHECK10-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
8032 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
8033 // CHECK10:       omp.inner.for.end:
8034 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8035 // CHECK10:       omp.loop.exit:
8036 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8037 // CHECK10-NEXT:    ret void
8038 //
8039 //
8040 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
8041 // CHECK10-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
8042 // CHECK10-NEXT:  entry:
8043 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8044 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8045 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8046 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8047 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8048 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8049 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8050 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8051 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8052 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8053 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
8054 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8055 // CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
8056 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8057 // CHECK10-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
8058 // CHECK10-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8059 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
8060 // CHECK10-NEXT:    ret void
8061 //
8062 //
8063 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
8064 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
8065 // CHECK10-NEXT:  entry:
8066 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8067 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8068 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8069 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8070 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8071 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8072 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8073 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8074 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8075 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8076 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8077 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8078 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8079 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8080 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8081 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8082 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8083 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8084 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8085 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8086 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8087 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8088 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8089 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8090 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8091 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
8092 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8093 // CHECK10:       cond.true:
8094 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8095 // CHECK10:       cond.false:
8096 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8097 // CHECK10-NEXT:    br label [[COND_END]]
8098 // CHECK10:       cond.end:
8099 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8100 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8101 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8102 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8103 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8104 // CHECK10:       omp.inner.for.cond:
8105 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8106 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8107 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8108 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8109 // CHECK10:       omp.inner.for.body:
8110 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8111 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
8112 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8113 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8114 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
8115 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
8116 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
8117 // CHECK10-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
8118 // CHECK10-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
8119 // CHECK10-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
8120 // CHECK10-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
8121 // CHECK10-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
8122 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8123 // CHECK10:       omp.body.continue:
8124 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8125 // CHECK10:       omp.inner.for.inc:
8126 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8127 // CHECK10-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
8128 // CHECK10-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
8129 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
8130 // CHECK10:       omp.inner.for.end:
8131 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8132 // CHECK10:       omp.loop.exit:
8133 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8134 // CHECK10-NEXT:    ret void
8135 //
8136 //
8137 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
8138 // CHECK10-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
8139 // CHECK10-NEXT:  entry:
8140 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8141 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
8142 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8143 // CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
8144 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
8145 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8146 // CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
8147 // CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
8148 // CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
8149 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8150 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8151 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
8152 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8153 // CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
8154 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8155 // CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
8156 // CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
8157 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8158 // CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
8159 // CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
8160 // CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
8161 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8162 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8163 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
8164 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8165 // CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
8166 // CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
8167 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8168 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
8169 // CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
8170 // CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
8171 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8172 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
8173 // CHECK10-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8174 // CHECK10-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
8175 // CHECK10-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
8176 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
8177 // CHECK10-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
8178 // CHECK10-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
8179 // CHECK10-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
8180 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
8181 // CHECK10-NEXT:    ret void
8182 //
8183 //
8184 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
8185 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
8186 // CHECK10-NEXT:  entry:
8187 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8188 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8189 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8190 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
8191 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8192 // CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
8193 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
8194 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8195 // CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
8196 // CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
8197 // CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
8198 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8199 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8200 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8201 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8202 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8203 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8204 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8205 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8206 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8207 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8208 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8209 // CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
8210 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8211 // CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
8212 // CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
8213 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8214 // CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
8215 // CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
8216 // CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
8217 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8218 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8219 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
8220 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8221 // CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
8222 // CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
8223 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8224 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
8225 // CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
8226 // CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
8227 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8228 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8229 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8230 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8231 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8232 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
8233 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8234 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8235 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
8236 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
8237 // CHECK10:       omp.dispatch.cond:
8238 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8239 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
8240 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8241 // CHECK10:       cond.true:
8242 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8243 // CHECK10:       cond.false:
8244 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8245 // CHECK10-NEXT:    br label [[COND_END]]
8246 // CHECK10:       cond.end:
8247 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8248 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8249 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8250 // CHECK10-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8251 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8252 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8253 // CHECK10-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8254 // CHECK10-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
8255 // CHECK10:       omp.dispatch.body:
8256 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8257 // CHECK10:       omp.inner.for.cond:
8258 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
8259 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
8260 // CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8261 // CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8262 // CHECK10:       omp.inner.for.body:
8263 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
8264 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8265 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8266 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
8267 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !13
8268 // CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
8269 // CHECK10-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !13
8270 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
8271 // CHECK10-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !13
8272 // CHECK10-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
8273 // CHECK10-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
8274 // CHECK10-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
8275 // CHECK10-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !13
8276 // CHECK10-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
8277 // CHECK10-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
8278 // CHECK10-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
8279 // CHECK10-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
8280 // CHECK10-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
8281 // CHECK10-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
8282 // CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
8283 // CHECK10-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
8284 // CHECK10-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
8285 // CHECK10-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
8286 // CHECK10-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
8287 // CHECK10-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
8288 // CHECK10-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
8289 // CHECK10-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
8290 // CHECK10-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
8291 // CHECK10-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
8292 // CHECK10-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
8293 // CHECK10-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
8294 // CHECK10-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !13
8295 // CHECK10-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
8296 // CHECK10-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !13
8297 // CHECK10-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
8298 // CHECK10-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !13
8299 // CHECK10-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
8300 // CHECK10-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
8301 // CHECK10-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
8302 // CHECK10-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !13
8303 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8304 // CHECK10:       omp.body.continue:
8305 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8306 // CHECK10:       omp.inner.for.inc:
8307 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
8308 // CHECK10-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
8309 // CHECK10-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
8310 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
8311 // CHECK10:       omp.inner.for.end:
8312 // CHECK10-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
8313 // CHECK10:       omp.dispatch.inc:
8314 // CHECK10-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8315 // CHECK10-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8316 // CHECK10-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
8317 // CHECK10-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
8318 // CHECK10-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8319 // CHECK10-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8320 // CHECK10-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
8321 // CHECK10-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
8322 // CHECK10-NEXT:    br label [[OMP_DISPATCH_COND]]
8323 // CHECK10:       omp.dispatch.end:
8324 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
8325 // CHECK10-NEXT:    ret void
8326 //
8327 //
8328 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
8329 // CHECK10-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
8330 // CHECK10-NEXT:  entry:
8331 // CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
8332 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8333 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8334 // CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
8335 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8336 // CHECK10-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
8337 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8338 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8339 // CHECK10-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
8340 // CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
8341 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8342 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8343 // CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
8344 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8345 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8346 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8347 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8348 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
8349 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8350 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
8351 // CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
8352 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
8353 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
8354 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
8355 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8356 // CHECK10-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
8357 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
8358 // CHECK10-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
8359 // CHECK10-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8360 // CHECK10-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
8361 // CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8362 // CHECK10-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
8363 // CHECK10-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
8364 // CHECK10-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
8365 // CHECK10-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
8366 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
8367 // CHECK10-NEXT:    ret void
8368 //
8369 //
8370 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
8371 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
8372 // CHECK10-NEXT:  entry:
8373 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8374 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8375 // CHECK10-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
8376 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8377 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8378 // CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
8379 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8380 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8381 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8382 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8383 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
8384 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
8385 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8386 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8387 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8388 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8389 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8390 // CHECK10-NEXT:    [[I8:%.*]] = alloca i32, align 4
8391 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8392 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8393 // CHECK10-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
8394 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8395 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8396 // CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
8397 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8398 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8399 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8400 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8401 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
8402 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8403 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
8404 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
8405 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
8406 // CHECK10-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
8407 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
8408 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8409 // CHECK10-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
8410 // CHECK10-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
8411 // CHECK10-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
8412 // CHECK10-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
8413 // CHECK10-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
8414 // CHECK10-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
8415 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8416 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
8417 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8418 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
8419 // CHECK10-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
8420 // CHECK10-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8421 // CHECK10:       omp.precond.then:
8422 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8423 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
8424 // CHECK10-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
8425 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8426 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8427 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8428 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8429 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8430 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8431 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
8432 // CHECK10-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
8433 // CHECK10-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8434 // CHECK10:       cond.true:
8435 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
8436 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8437 // CHECK10:       cond.false:
8438 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8439 // CHECK10-NEXT:    br label [[COND_END]]
8440 // CHECK10:       cond.end:
8441 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8442 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8443 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8444 // CHECK10-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8445 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8446 // CHECK10:       omp.inner.for.cond:
8447 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8448 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8449 // CHECK10-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
8450 // CHECK10-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
8451 // CHECK10-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8452 // CHECK10:       omp.inner.for.body:
8453 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8454 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8455 // CHECK10-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
8456 // CHECK10-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
8457 // CHECK10-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
8458 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
8459 // CHECK10-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
8460 // CHECK10-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
8461 // CHECK10-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
8462 // CHECK10-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
8463 // CHECK10-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
8464 // CHECK10-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
8465 // CHECK10-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
8466 // CHECK10-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
8467 // CHECK10-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
8468 // CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
8469 // CHECK10-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
8470 // CHECK10-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
8471 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
8472 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
8473 // CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
8474 // CHECK10-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
8475 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8476 // CHECK10:       omp.body.continue:
8477 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8478 // CHECK10:       omp.inner.for.inc:
8479 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8480 // CHECK10-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
8481 // CHECK10-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
8482 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
8483 // CHECK10:       omp.inner.for.end:
8484 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8485 // CHECK10:       omp.loop.exit:
8486 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8487 // CHECK10-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
8488 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
8489 // CHECK10-NEXT:    br label [[OMP_PRECOND_END]]
8490 // CHECK10:       omp.precond.end:
8491 // CHECK10-NEXT:    ret void
8492 //
8493 //
8494 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
8495 // CHECK10-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
8496 // CHECK10-NEXT:  entry:
8497 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8498 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
8499 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8500 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8501 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
8502 // CHECK10-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
8503 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8504 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
8505 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8506 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8507 // CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
8508 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8509 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
8510 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8511 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8512 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
8513 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
8514 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
8515 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
8516 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
8517 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
8518 // CHECK10-NEXT:    ret void
8519 //
8520 //
8521 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
8522 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
8523 // CHECK10-NEXT:  entry:
8524 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8525 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8526 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8527 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
8528 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8529 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8530 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
8531 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8532 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8533 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8534 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8535 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8536 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8537 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8538 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8539 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8540 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8541 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
8542 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8543 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8544 // CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
8545 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8546 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
8547 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8548 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8549 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
8550 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8551 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8552 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8553 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8554 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8555 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
8556 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8557 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8558 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
8559 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8560 // CHECK10:       cond.true:
8561 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8562 // CHECK10:       cond.false:
8563 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8564 // CHECK10-NEXT:    br label [[COND_END]]
8565 // CHECK10:       cond.end:
8566 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
8567 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8568 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8569 // CHECK10-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
8570 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8571 // CHECK10:       omp.inner.for.cond:
8572 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8573 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8574 // CHECK10-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
8575 // CHECK10-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8576 // CHECK10:       omp.inner.for.body:
8577 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8578 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
8579 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8580 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8581 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
8582 // CHECK10-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
8583 // CHECK10-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
8584 // CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
8585 // CHECK10-NEXT:    store double [[ADD5]], double* [[A]], align 8
8586 // CHECK10-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
8587 // CHECK10-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
8588 // CHECK10-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
8589 // CHECK10-NEXT:    store double [[INC]], double* [[A6]], align 8
8590 // CHECK10-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
8591 // CHECK10-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
8592 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
8593 // CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
8594 // CHECK10-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
8595 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8596 // CHECK10:       omp.body.continue:
8597 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8598 // CHECK10:       omp.inner.for.inc:
8599 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8600 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
8601 // CHECK10-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
8602 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
8603 // CHECK10:       omp.inner.for.end:
8604 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8605 // CHECK10:       omp.loop.exit:
8606 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
8607 // CHECK10-NEXT:    ret void
8608 //
8609 //
8610 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
8611 // CHECK10-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
8612 // CHECK10-NEXT:  entry:
8613 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8614 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8615 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8616 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8617 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8618 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8619 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8620 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8621 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8622 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8623 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8624 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
8625 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8626 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
8627 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
8628 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
8629 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8630 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
8631 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8632 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
8633 // CHECK10-NEXT:    ret void
8634 //
8635 //
8636 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
8637 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
8638 // CHECK10-NEXT:  entry:
8639 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8640 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8641 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8642 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8643 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
8644 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8645 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8646 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8647 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8648 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8649 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8650 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8651 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8652 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8653 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8654 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8655 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
8656 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8657 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8658 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
8659 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8660 // CHECK10-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8661 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8662 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8663 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8664 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8665 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8666 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8667 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
8668 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8669 // CHECK10:       cond.true:
8670 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8671 // CHECK10:       cond.false:
8672 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8673 // CHECK10-NEXT:    br label [[COND_END]]
8674 // CHECK10:       cond.end:
8675 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8676 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8677 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8678 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
8679 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8680 // CHECK10:       omp.inner.for.cond:
8681 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8682 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8683 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8684 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8685 // CHECK10:       omp.inner.for.body:
8686 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8687 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
8688 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8689 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8690 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
8691 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
8692 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
8693 // CHECK10-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
8694 // CHECK10-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
8695 // CHECK10-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
8696 // CHECK10-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
8697 // CHECK10-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
8698 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
8699 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
8700 // CHECK10-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
8701 // CHECK10-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
8702 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8703 // CHECK10:       omp.body.continue:
8704 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8705 // CHECK10:       omp.inner.for.inc:
8706 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8707 // CHECK10-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
8708 // CHECK10-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
8709 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
8710 // CHECK10:       omp.inner.for.end:
8711 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8712 // CHECK10:       omp.loop.exit:
8713 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
8714 // CHECK10-NEXT:    ret void
8715 //
8716 //
8717 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
8718 // CHECK11-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
8719 // CHECK11-NEXT:  entry:
8720 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8721 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
8722 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
8723 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
8724 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
8725 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8726 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8727 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
8728 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8729 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8730 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
8731 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
8732 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
8733 // CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
8734 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
8735 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
8736 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
8737 // CHECK11-NEXT:    ret void
8738 //
8739 //
8740 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
8741 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
8742 // CHECK11-NEXT:  entry:
8743 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8744 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8745 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8746 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8747 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8748 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8749 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8750 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8751 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8752 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
8753 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8754 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8755 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8756 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8757 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8758 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8759 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8760 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8761 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8762 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8763 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8764 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8765 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
8766 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8767 // CHECK11:       cond.true:
8768 // CHECK11-NEXT:    br label [[COND_END:%.*]]
8769 // CHECK11:       cond.false:
8770 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8771 // CHECK11-NEXT:    br label [[COND_END]]
8772 // CHECK11:       cond.end:
8773 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8774 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8775 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8776 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8777 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8778 // CHECK11:       omp.inner.for.cond:
8779 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8780 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8781 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8782 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8783 // CHECK11:       omp.inner.for.body:
8784 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8785 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
8786 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8787 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8788 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8789 // CHECK11:       omp.body.continue:
8790 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8791 // CHECK11:       omp.inner.for.inc:
8792 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8793 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
8794 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
8795 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
8796 // CHECK11:       omp.inner.for.end:
8797 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8798 // CHECK11:       omp.loop.exit:
8799 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8800 // CHECK11-NEXT:    ret void
8801 //
8802 //
8803 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
8804 // CHECK11-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
8805 // CHECK11-NEXT:  entry:
8806 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8807 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
8808 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8809 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8810 // CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
8811 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
8812 // CHECK11-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
8813 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
8814 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
8815 // CHECK11-NEXT:    ret void
8816 //
8817 //
8818 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
8819 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
8820 // CHECK11-NEXT:  entry:
8821 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8822 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8823 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8824 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8825 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8826 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8827 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8828 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8829 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8830 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
8831 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8832 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8833 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8834 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8835 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8836 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8837 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8838 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8839 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8840 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8841 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8842 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8843 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
8844 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8845 // CHECK11:       cond.true:
8846 // CHECK11-NEXT:    br label [[COND_END:%.*]]
8847 // CHECK11:       cond.false:
8848 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8849 // CHECK11-NEXT:    br label [[COND_END]]
8850 // CHECK11:       cond.end:
8851 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8852 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8853 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8854 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8855 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8856 // CHECK11:       omp.inner.for.cond:
8857 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8858 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8859 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8860 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8861 // CHECK11:       omp.inner.for.body:
8862 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8863 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
8864 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8865 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8866 // CHECK11-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
8867 // CHECK11-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
8868 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
8869 // CHECK11-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
8870 // CHECK11-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
8871 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8872 // CHECK11:       omp.body.continue:
8873 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8874 // CHECK11:       omp.inner.for.inc:
8875 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8876 // CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
8877 // CHECK11-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
8878 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
8879 // CHECK11:       omp.inner.for.end:
8880 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8881 // CHECK11:       omp.loop.exit:
8882 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8883 // CHECK11-NEXT:    ret void
8884 //
8885 //
8886 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
8887 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
8888 // CHECK11-NEXT:  entry:
8889 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8890 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8891 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
8892 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
8893 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8894 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8895 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8896 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
8897 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
8898 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
8899 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
8900 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
8901 // CHECK11-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
8902 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
8903 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
8904 // CHECK11-NEXT:    ret void
8905 //
8906 //
8907 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
8908 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
8909 // CHECK11-NEXT:  entry:
8910 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8911 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8912 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8913 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
8914 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8915 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8916 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8917 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8918 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8919 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8920 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
8921 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8922 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8923 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8924 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
8925 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
8926 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8927 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
8928 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8929 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8930 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8931 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8932 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8933 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8934 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
8935 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8936 // CHECK11:       cond.true:
8937 // CHECK11-NEXT:    br label [[COND_END:%.*]]
8938 // CHECK11:       cond.false:
8939 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8940 // CHECK11-NEXT:    br label [[COND_END]]
8941 // CHECK11:       cond.end:
8942 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8943 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8944 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8945 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8946 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8947 // CHECK11:       omp.inner.for.cond:
8948 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8949 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8950 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8951 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8952 // CHECK11:       omp.inner.for.body:
8953 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8954 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
8955 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8956 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
8957 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
8958 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
8959 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
8960 // CHECK11-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
8961 // CHECK11-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
8962 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
8963 // CHECK11-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
8964 // CHECK11-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
8965 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8966 // CHECK11:       omp.body.continue:
8967 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8968 // CHECK11:       omp.inner.for.inc:
8969 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8970 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
8971 // CHECK11-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
8972 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
8973 // CHECK11:       omp.inner.for.end:
8974 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8975 // CHECK11:       omp.loop.exit:
8976 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8977 // CHECK11-NEXT:    ret void
8978 //
8979 //
8980 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
8981 // CHECK11-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
8982 // CHECK11-NEXT:  entry:
8983 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
8984 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
8985 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
8986 // CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
8987 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
8988 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
8989 // CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
8990 // CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
8991 // CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
8992 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
8993 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
8994 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
8995 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
8996 // CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
8997 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8998 // CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
8999 // CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9000 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9001 // CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9002 // CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9003 // CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9004 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9005 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9006 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9007 // CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9008 // CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9009 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9010 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9011 // CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9012 // CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9013 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
9014 // CHECK11-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
9015 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
9016 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9017 // CHECK11-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9018 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9019 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
9020 // CHECK11-NEXT:    ret void
9021 //
9022 //
9023 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
9024 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
9025 // CHECK11-NEXT:  entry:
9026 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9027 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9028 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9029 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9030 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9031 // CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9032 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9033 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9034 // CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9035 // CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9036 // CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9037 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9038 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9039 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9040 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9041 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9042 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9043 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9044 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
9045 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9046 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9047 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9048 // CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9049 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9050 // CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9051 // CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9052 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9053 // CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9054 // CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9055 // CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9056 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9057 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9058 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9059 // CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9060 // CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9061 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9062 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9063 // CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9064 // CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9065 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9066 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9067 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9068 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9069 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9070 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9071 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
9072 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
9073 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
9074 // CHECK11:       omp.dispatch.cond:
9075 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9076 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
9077 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9078 // CHECK11:       cond.true:
9079 // CHECK11-NEXT:    br label [[COND_END:%.*]]
9080 // CHECK11:       cond.false:
9081 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9082 // CHECK11-NEXT:    br label [[COND_END]]
9083 // CHECK11:       cond.end:
9084 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
9085 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9086 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9087 // CHECK11-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
9088 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9089 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9090 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
9091 // CHECK11-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9092 // CHECK11:       omp.dispatch.body:
9093 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9094 // CHECK11:       omp.inner.for.cond:
9095 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9096 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
9097 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
9098 // CHECK11-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9099 // CHECK11:       omp.inner.for.body:
9100 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9101 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
9102 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9103 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
9104 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !14
9105 // CHECK11-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
9106 // CHECK11-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !14
9107 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
9108 // CHECK11-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
9109 // CHECK11-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
9110 // CHECK11-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
9111 // CHECK11-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
9112 // CHECK11-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !14
9113 // CHECK11-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
9114 // CHECK11-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
9115 // CHECK11-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
9116 // CHECK11-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
9117 // CHECK11-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
9118 // CHECK11-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
9119 // CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
9120 // CHECK11-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
9121 // CHECK11-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
9122 // CHECK11-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
9123 // CHECK11-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
9124 // CHECK11-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
9125 // CHECK11-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
9126 // CHECK11-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
9127 // CHECK11-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
9128 // CHECK11-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
9129 // CHECK11-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
9130 // CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
9131 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !14
9132 // CHECK11-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
9133 // CHECK11-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !14
9134 // CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
9135 // CHECK11-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !14
9136 // CHECK11-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
9137 // CHECK11-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
9138 // CHECK11-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
9139 // CHECK11-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !14
9140 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9141 // CHECK11:       omp.body.continue:
9142 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9143 // CHECK11:       omp.inner.for.inc:
9144 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9145 // CHECK11-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
9146 // CHECK11-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9147 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
9148 // CHECK11:       omp.inner.for.end:
9149 // CHECK11-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
9150 // CHECK11:       omp.dispatch.inc:
9151 // CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9152 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9153 // CHECK11-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
9154 // CHECK11-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
9155 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9156 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9157 // CHECK11-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
9158 // CHECK11-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
9159 // CHECK11-NEXT:    br label [[OMP_DISPATCH_COND]]
9160 // CHECK11:       omp.dispatch.end:
9161 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
9162 // CHECK11-NEXT:    ret void
9163 //
9164 //
9165 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
9166 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9167 // CHECK11-NEXT:  entry:
9168 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9169 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9170 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9171 // CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
9172 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9173 // CHECK11-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
9174 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9175 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9176 // CHECK11-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
9177 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9178 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9179 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9180 // CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
9181 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9182 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9183 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
9184 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9185 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
9186 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
9187 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
9188 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
9189 // CHECK11-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
9190 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
9191 // CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
9192 // CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9193 // CHECK11-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
9194 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9195 // CHECK11-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
9196 // CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
9197 // CHECK11-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
9198 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
9199 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
9200 // CHECK11-NEXT:    ret void
9201 //
9202 //
9203 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
9204 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9205 // CHECK11-NEXT:  entry:
9206 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9207 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9208 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9209 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9210 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9211 // CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
9212 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9213 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9214 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9215 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9216 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
9217 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
9218 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
9219 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9220 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9221 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9222 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9223 // CHECK11-NEXT:    [[I6:%.*]] = alloca i32, align 4
9224 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9225 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9226 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9227 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9228 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9229 // CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
9230 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9231 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9232 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
9233 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9234 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
9235 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
9236 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
9237 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
9238 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
9239 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9240 // CHECK11-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
9241 // CHECK11-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
9242 // CHECK11-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
9243 // CHECK11-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
9244 // CHECK11-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
9245 // CHECK11-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
9246 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9247 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
9248 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9249 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
9250 // CHECK11-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
9251 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
9252 // CHECK11:       omp.precond.then:
9253 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9254 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
9255 // CHECK11-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
9256 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9257 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9258 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9259 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
9260 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9261 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9262 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
9263 // CHECK11-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
9264 // CHECK11-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9265 // CHECK11:       cond.true:
9266 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
9267 // CHECK11-NEXT:    br label [[COND_END:%.*]]
9268 // CHECK11:       cond.false:
9269 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9270 // CHECK11-NEXT:    br label [[COND_END]]
9271 // CHECK11:       cond.end:
9272 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
9273 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9274 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9275 // CHECK11-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
9276 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9277 // CHECK11:       omp.inner.for.cond:
9278 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9279 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9280 // CHECK11-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
9281 // CHECK11-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
9282 // CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9283 // CHECK11:       omp.inner.for.body:
9284 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9285 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9286 // CHECK11-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
9287 // CHECK11-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
9288 // CHECK11-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
9289 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
9290 // CHECK11-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
9291 // CHECK11-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
9292 // CHECK11-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
9293 // CHECK11-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
9294 // CHECK11-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
9295 // CHECK11-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
9296 // CHECK11-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
9297 // CHECK11-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
9298 // CHECK11-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
9299 // CHECK11-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
9300 // CHECK11-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
9301 // CHECK11-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
9302 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
9303 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9304 // CHECK11-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
9305 // CHECK11-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
9306 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9307 // CHECK11:       omp.body.continue:
9308 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9309 // CHECK11:       omp.inner.for.inc:
9310 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9311 // CHECK11-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
9312 // CHECK11-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
9313 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
9314 // CHECK11:       omp.inner.for.end:
9315 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9316 // CHECK11:       omp.loop.exit:
9317 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9318 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
9319 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
9320 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
9321 // CHECK11:       omp.precond.end:
9322 // CHECK11-NEXT:    ret void
9323 //
9324 //
9325 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
9326 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
9327 // CHECK11-NEXT:  entry:
9328 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
9329 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
9330 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9331 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9332 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
9333 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
9334 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
9335 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
9336 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9337 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9338 // CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
9339 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
9340 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9341 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9342 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
9343 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
9344 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
9345 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
9346 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
9347 // CHECK11-NEXT:    ret void
9348 //
9349 //
9350 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
9351 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
9352 // CHECK11-NEXT:  entry:
9353 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9354 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9355 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
9356 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
9357 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9358 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9359 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
9360 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9361 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9362 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9363 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9364 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9365 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9366 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
9367 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9368 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9369 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
9370 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
9371 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9372 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9373 // CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
9374 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
9375 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9376 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9377 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
9378 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9379 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9380 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9381 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9382 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9383 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
9384 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9385 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9386 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
9387 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9388 // CHECK11:       cond.true:
9389 // CHECK11-NEXT:    br label [[COND_END:%.*]]
9390 // CHECK11:       cond.false:
9391 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9392 // CHECK11-NEXT:    br label [[COND_END]]
9393 // CHECK11:       cond.end:
9394 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
9395 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9396 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9397 // CHECK11-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
9398 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9399 // CHECK11:       omp.inner.for.cond:
9400 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9401 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9402 // CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
9403 // CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9404 // CHECK11:       omp.inner.for.body:
9405 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9406 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
9407 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9408 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
9409 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
9410 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
9411 // CHECK11-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
9412 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
9413 // CHECK11-NEXT:    store double [[ADD4]], double* [[A]], align 4
9414 // CHECK11-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
9415 // CHECK11-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
9416 // CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
9417 // CHECK11-NEXT:    store double [[INC]], double* [[A5]], align 4
9418 // CHECK11-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
9419 // CHECK11-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
9420 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
9421 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
9422 // CHECK11-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
9423 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9424 // CHECK11:       omp.body.continue:
9425 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9426 // CHECK11:       omp.inner.for.inc:
9427 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9428 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
9429 // CHECK11-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
9430 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
9431 // CHECK11:       omp.inner.for.end:
9432 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9433 // CHECK11:       omp.loop.exit:
9434 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
9435 // CHECK11-NEXT:    ret void
9436 //
9437 //
9438 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
9439 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9440 // CHECK11-NEXT:  entry:
9441 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9442 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9443 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9444 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9445 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9446 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9447 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9448 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9449 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9450 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9451 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
9452 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
9453 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
9454 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
9455 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9456 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
9457 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9458 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
9459 // CHECK11-NEXT:    ret void
9460 //
9461 //
9462 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
9463 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9464 // CHECK11-NEXT:  entry:
9465 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9466 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9467 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9468 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9469 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9470 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9471 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9472 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9473 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9474 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9475 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9476 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
9477 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9478 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9479 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9480 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9481 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
9482 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9483 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
9484 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9485 // CHECK11-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9486 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9487 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9488 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9489 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9490 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9491 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9492 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9493 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9494 // CHECK11:       cond.true:
9495 // CHECK11-NEXT:    br label [[COND_END:%.*]]
9496 // CHECK11:       cond.false:
9497 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9498 // CHECK11-NEXT:    br label [[COND_END]]
9499 // CHECK11:       cond.end:
9500 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9501 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9502 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9503 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9504 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9505 // CHECK11:       omp.inner.for.cond:
9506 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9507 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9508 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9509 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9510 // CHECK11:       omp.inner.for.body:
9511 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9512 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
9513 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9514 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
9515 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
9516 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
9517 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
9518 // CHECK11-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
9519 // CHECK11-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
9520 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
9521 // CHECK11-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
9522 // CHECK11-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
9523 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
9524 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9525 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
9526 // CHECK11-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
9527 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9528 // CHECK11:       omp.body.continue:
9529 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9530 // CHECK11:       omp.inner.for.inc:
9531 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9532 // CHECK11-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
9533 // CHECK11-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
9534 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
9535 // CHECK11:       omp.inner.for.end:
9536 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9537 // CHECK11:       omp.loop.exit:
9538 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9539 // CHECK11-NEXT:    ret void
9540 //
9541 //
9542 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
9543 // CHECK12-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
9544 // CHECK12-NEXT:  entry:
9545 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9546 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9547 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
9548 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9549 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
9550 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9551 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9552 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
9553 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9554 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9555 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
9556 // CHECK12-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
9557 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
9558 // CHECK12-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9559 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
9560 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9561 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
9562 // CHECK12-NEXT:    ret void
9563 //
9564 //
9565 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
9566 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9567 // CHECK12-NEXT:  entry:
9568 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9569 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9570 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9571 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9572 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9573 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9574 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9575 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9576 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9577 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
9578 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9579 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9580 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9581 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9582 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9583 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9584 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9585 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9586 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9587 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9588 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9589 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9590 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
9591 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9592 // CHECK12:       cond.true:
9593 // CHECK12-NEXT:    br label [[COND_END:%.*]]
9594 // CHECK12:       cond.false:
9595 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9596 // CHECK12-NEXT:    br label [[COND_END]]
9597 // CHECK12:       cond.end:
9598 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9599 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9600 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9601 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9602 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9603 // CHECK12:       omp.inner.for.cond:
9604 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9605 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9606 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9607 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9608 // CHECK12:       omp.inner.for.body:
9609 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9610 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
9611 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9612 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
9613 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9614 // CHECK12:       omp.body.continue:
9615 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9616 // CHECK12:       omp.inner.for.inc:
9617 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9618 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
9619 // CHECK12-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
9620 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
9621 // CHECK12:       omp.inner.for.end:
9622 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9623 // CHECK12:       omp.loop.exit:
9624 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9625 // CHECK12-NEXT:    ret void
9626 //
9627 //
9628 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
9629 // CHECK12-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9630 // CHECK12-NEXT:  entry:
9631 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9632 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9633 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9634 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9635 // CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
9636 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9637 // CHECK12-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
9638 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9639 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
9640 // CHECK12-NEXT:    ret void
9641 //
9642 //
9643 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
9644 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9645 // CHECK12-NEXT:  entry:
9646 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9647 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9648 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9649 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9650 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9651 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9652 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9653 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9654 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9655 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
9656 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9657 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9658 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9659 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9660 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9661 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9662 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9663 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9664 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9665 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9666 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9667 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9668 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
9669 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9670 // CHECK12:       cond.true:
9671 // CHECK12-NEXT:    br label [[COND_END:%.*]]
9672 // CHECK12:       cond.false:
9673 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9674 // CHECK12-NEXT:    br label [[COND_END]]
9675 // CHECK12:       cond.end:
9676 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9677 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9678 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9679 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9680 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9681 // CHECK12:       omp.inner.for.cond:
9682 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9683 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9684 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9685 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9686 // CHECK12:       omp.inner.for.body:
9687 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9688 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
9689 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9690 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
9691 // CHECK12-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
9692 // CHECK12-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
9693 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
9694 // CHECK12-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
9695 // CHECK12-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
9696 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9697 // CHECK12:       omp.body.continue:
9698 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9699 // CHECK12:       omp.inner.for.inc:
9700 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9701 // CHECK12-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
9702 // CHECK12-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
9703 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
9704 // CHECK12:       omp.inner.for.end:
9705 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9706 // CHECK12:       omp.loop.exit:
9707 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9708 // CHECK12-NEXT:    ret void
9709 //
9710 //
9711 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
9712 // CHECK12-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9713 // CHECK12-NEXT:  entry:
9714 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9715 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9716 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9717 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9718 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9719 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9720 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9721 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
9722 // CHECK12-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
9723 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
9724 // CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
9725 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9726 // CHECK12-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
9727 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9728 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
9729 // CHECK12-NEXT:    ret void
9730 //
9731 //
9732 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
9733 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9734 // CHECK12-NEXT:  entry:
9735 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9736 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9737 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9738 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9739 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9740 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9741 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9742 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9743 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9744 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9745 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
9746 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9747 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9748 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9749 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9750 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9751 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9752 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9753 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9754 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9755 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9756 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9757 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9758 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9759 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
9760 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9761 // CHECK12:       cond.true:
9762 // CHECK12-NEXT:    br label [[COND_END:%.*]]
9763 // CHECK12:       cond.false:
9764 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9765 // CHECK12-NEXT:    br label [[COND_END]]
9766 // CHECK12:       cond.end:
9767 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9768 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9769 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9770 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9771 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9772 // CHECK12:       omp.inner.for.cond:
9773 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9774 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9775 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9776 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9777 // CHECK12:       omp.inner.for.body:
9778 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9779 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
9780 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9781 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
9782 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
9783 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
9784 // CHECK12-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
9785 // CHECK12-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
9786 // CHECK12-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
9787 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
9788 // CHECK12-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
9789 // CHECK12-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
9790 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9791 // CHECK12:       omp.body.continue:
9792 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9793 // CHECK12:       omp.inner.for.inc:
9794 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9795 // CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
9796 // CHECK12-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
9797 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
9798 // CHECK12:       omp.inner.for.end:
9799 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9800 // CHECK12:       omp.loop.exit:
9801 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9802 // CHECK12-NEXT:    ret void
9803 //
9804 //
9805 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
9806 // CHECK12-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
9807 // CHECK12-NEXT:  entry:
9808 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9809 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9810 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9811 // CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9812 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9813 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9814 // CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9815 // CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9816 // CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9817 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9818 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9819 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
9820 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9821 // CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9822 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9823 // CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9824 // CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9825 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9826 // CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9827 // CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9828 // CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9829 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9830 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9831 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9832 // CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9833 // CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9834 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9835 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9836 // CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9837 // CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9838 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
9839 // CHECK12-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
9840 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
9841 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9842 // CHECK12-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9843 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9844 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
9845 // CHECK12-NEXT:    ret void
9846 //
9847 //
9848 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
9849 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
9850 // CHECK12-NEXT:  entry:
9851 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9852 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9853 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9854 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9855 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9856 // CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9857 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9858 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9859 // CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9860 // CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9861 // CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9862 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9863 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9864 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9865 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9866 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9867 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9868 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9869 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
9870 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9871 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9872 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9873 // CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9874 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9875 // CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9876 // CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9877 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9878 // CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9879 // CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9880 // CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9881 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9882 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9883 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9884 // CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9885 // CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9886 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9887 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9888 // CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9889 // CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9890 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9891 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
9892 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9893 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9894 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9895 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9896 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
9897 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
9898 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
9899 // CHECK12:       omp.dispatch.cond:
9900 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9901 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
9902 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9903 // CHECK12:       cond.true:
9904 // CHECK12-NEXT:    br label [[COND_END:%.*]]
9905 // CHECK12:       cond.false:
9906 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9907 // CHECK12-NEXT:    br label [[COND_END]]
9908 // CHECK12:       cond.end:
9909 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
9910 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9911 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9912 // CHECK12-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
9913 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9914 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9915 // CHECK12-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
9916 // CHECK12-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9917 // CHECK12:       omp.dispatch.body:
9918 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9919 // CHECK12:       omp.inner.for.cond:
9920 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9921 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
9922 // CHECK12-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
9923 // CHECK12-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9924 // CHECK12:       omp.inner.for.body:
9925 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9926 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
9927 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9928 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
9929 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !14
9930 // CHECK12-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
9931 // CHECK12-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !14
9932 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
9933 // CHECK12-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
9934 // CHECK12-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
9935 // CHECK12-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
9936 // CHECK12-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
9937 // CHECK12-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !14
9938 // CHECK12-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
9939 // CHECK12-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
9940 // CHECK12-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
9941 // CHECK12-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
9942 // CHECK12-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
9943 // CHECK12-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
9944 // CHECK12-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
9945 // CHECK12-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
9946 // CHECK12-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
9947 // CHECK12-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
9948 // CHECK12-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
9949 // CHECK12-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
9950 // CHECK12-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
9951 // CHECK12-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
9952 // CHECK12-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
9953 // CHECK12-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
9954 // CHECK12-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
9955 // CHECK12-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
9956 // CHECK12-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !14
9957 // CHECK12-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
9958 // CHECK12-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !14
9959 // CHECK12-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
9960 // CHECK12-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !14
9961 // CHECK12-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
9962 // CHECK12-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
9963 // CHECK12-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
9964 // CHECK12-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !14
9965 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9966 // CHECK12:       omp.body.continue:
9967 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9968 // CHECK12:       omp.inner.for.inc:
9969 // CHECK12-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9970 // CHECK12-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
9971 // CHECK12-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
9972 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
9973 // CHECK12:       omp.inner.for.end:
9974 // CHECK12-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
9975 // CHECK12:       omp.dispatch.inc:
9976 // CHECK12-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9977 // CHECK12-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9978 // CHECK12-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
9979 // CHECK12-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
9980 // CHECK12-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9981 // CHECK12-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9982 // CHECK12-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
9983 // CHECK12-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
9984 // CHECK12-NEXT:    br label [[OMP_DISPATCH_COND]]
9985 // CHECK12:       omp.dispatch.end:
9986 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
9987 // CHECK12-NEXT:    ret void
9988 //
9989 //
9990 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
9991 // CHECK12-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9992 // CHECK12-NEXT:  entry:
9993 // CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9994 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9995 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9996 // CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
9997 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
9998 // CHECK12-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
9999 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10000 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10001 // CHECK12-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
10002 // CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10003 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10004 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10005 // CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
10006 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10007 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10008 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
10009 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10010 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
10011 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
10012 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
10013 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
10014 // CHECK12-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
10015 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
10016 // CHECK12-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
10017 // CHECK12-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10018 // CHECK12-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
10019 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10020 // CHECK12-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
10021 // CHECK12-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
10022 // CHECK12-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
10023 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
10024 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
10025 // CHECK12-NEXT:    ret void
10026 //
10027 //
10028 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
10029 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10030 // CHECK12-NEXT:  entry:
10031 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10032 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10033 // CHECK12-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10034 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10035 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10036 // CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
10037 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10038 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10039 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10040 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10041 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
10042 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
10043 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
10044 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10045 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10046 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10047 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10048 // CHECK12-NEXT:    [[I6:%.*]] = alloca i32, align 4
10049 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10050 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10051 // CHECK12-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10052 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10053 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10054 // CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
10055 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10056 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10057 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
10058 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10059 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
10060 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
10061 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
10062 // CHECK12-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
10063 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10064 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10065 // CHECK12-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
10066 // CHECK12-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
10067 // CHECK12-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
10068 // CHECK12-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
10069 // CHECK12-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
10070 // CHECK12-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
10071 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10072 // CHECK12-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
10073 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10074 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10075 // CHECK12-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
10076 // CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10077 // CHECK12:       omp.precond.then:
10078 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10079 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
10080 // CHECK12-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
10081 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10082 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10083 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10084 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
10085 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10086 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10087 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
10088 // CHECK12-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
10089 // CHECK12-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10090 // CHECK12:       cond.true:
10091 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
10092 // CHECK12-NEXT:    br label [[COND_END:%.*]]
10093 // CHECK12:       cond.false:
10094 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10095 // CHECK12-NEXT:    br label [[COND_END]]
10096 // CHECK12:       cond.end:
10097 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
10098 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10099 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10100 // CHECK12-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
10101 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10102 // CHECK12:       omp.inner.for.cond:
10103 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10104 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10105 // CHECK12-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
10106 // CHECK12-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
10107 // CHECK12-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10108 // CHECK12:       omp.inner.for.body:
10109 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10110 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10111 // CHECK12-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
10112 // CHECK12-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
10113 // CHECK12-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
10114 // CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
10115 // CHECK12-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
10116 // CHECK12-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
10117 // CHECK12-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
10118 // CHECK12-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
10119 // CHECK12-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
10120 // CHECK12-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
10121 // CHECK12-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
10122 // CHECK12-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
10123 // CHECK12-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
10124 // CHECK12-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
10125 // CHECK12-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
10126 // CHECK12-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
10127 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
10128 // CHECK12-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10129 // CHECK12-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
10130 // CHECK12-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
10131 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10132 // CHECK12:       omp.body.continue:
10133 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10134 // CHECK12:       omp.inner.for.inc:
10135 // CHECK12-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10136 // CHECK12-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
10137 // CHECK12-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
10138 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
10139 // CHECK12:       omp.inner.for.end:
10140 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10141 // CHECK12:       omp.loop.exit:
10142 // CHECK12-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10143 // CHECK12-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
10144 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
10145 // CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
10146 // CHECK12:       omp.precond.end:
10147 // CHECK12-NEXT:    ret void
10148 //
10149 //
10150 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
10151 // CHECK12-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
10152 // CHECK12-NEXT:  entry:
10153 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10154 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
10155 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
10156 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
10157 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
10158 // CHECK12-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
10159 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10160 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
10161 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
10162 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
10163 // CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
10164 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10165 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
10166 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
10167 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
10168 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
10169 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
10170 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
10171 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
10172 // CHECK12-NEXT:    ret void
10173 //
10174 //
10175 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
10176 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
10177 // CHECK12-NEXT:  entry:
10178 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10179 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10180 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10181 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
10182 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
10183 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
10184 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
10185 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10186 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10187 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10188 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10189 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10190 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10191 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
10192 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10193 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10194 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10195 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
10196 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
10197 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
10198 // CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
10199 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10200 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
10201 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
10202 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
10203 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10204 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
10205 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10206 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10207 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10208 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
10209 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10210 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10211 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
10212 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10213 // CHECK12:       cond.true:
10214 // CHECK12-NEXT:    br label [[COND_END:%.*]]
10215 // CHECK12:       cond.false:
10216 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10217 // CHECK12-NEXT:    br label [[COND_END]]
10218 // CHECK12:       cond.end:
10219 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
10220 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10221 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10222 // CHECK12-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
10223 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10224 // CHECK12:       omp.inner.for.cond:
10225 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10226 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10227 // CHECK12-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
10228 // CHECK12-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10229 // CHECK12:       omp.inner.for.body:
10230 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10231 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
10232 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10233 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
10234 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
10235 // CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
10236 // CHECK12-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
10237 // CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
10238 // CHECK12-NEXT:    store double [[ADD4]], double* [[A]], align 4
10239 // CHECK12-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
10240 // CHECK12-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
10241 // CHECK12-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
10242 // CHECK12-NEXT:    store double [[INC]], double* [[A5]], align 4
10243 // CHECK12-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
10244 // CHECK12-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
10245 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
10246 // CHECK12-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
10247 // CHECK12-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
10248 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10249 // CHECK12:       omp.body.continue:
10250 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10251 // CHECK12:       omp.inner.for.inc:
10252 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10253 // CHECK12-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
10254 // CHECK12-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
10255 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
10256 // CHECK12:       omp.inner.for.end:
10257 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10258 // CHECK12:       omp.loop.exit:
10259 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
10260 // CHECK12-NEXT:    ret void
10261 //
10262 //
10263 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
10264 // CHECK12-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10265 // CHECK12-NEXT:  entry:
10266 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10267 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10268 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10269 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10270 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10271 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10272 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10273 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10274 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10275 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10276 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
10277 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
10278 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
10279 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
10280 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10281 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
10282 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10283 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
10284 // CHECK12-NEXT:    ret void
10285 //
10286 //
10287 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
10288 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10289 // CHECK12-NEXT:  entry:
10290 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10291 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10292 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10293 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10294 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10295 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10296 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10297 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10298 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10299 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10300 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10301 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
10302 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10303 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10304 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10305 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10306 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10307 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10308 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10309 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10310 // CHECK12-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
10311 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10312 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10313 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10314 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10315 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10316 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10317 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
10318 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10319 // CHECK12:       cond.true:
10320 // CHECK12-NEXT:    br label [[COND_END:%.*]]
10321 // CHECK12:       cond.false:
10322 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10323 // CHECK12-NEXT:    br label [[COND_END]]
10324 // CHECK12:       cond.end:
10325 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
10326 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10327 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10328 // CHECK12-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
10329 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10330 // CHECK12:       omp.inner.for.cond:
10331 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10332 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10333 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
10334 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10335 // CHECK12:       omp.inner.for.body:
10336 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10337 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
10338 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10339 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
10340 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
10341 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
10342 // CHECK12-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
10343 // CHECK12-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
10344 // CHECK12-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
10345 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
10346 // CHECK12-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
10347 // CHECK12-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
10348 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
10349 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10350 // CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
10351 // CHECK12-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
10352 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10353 // CHECK12:       omp.body.continue:
10354 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10355 // CHECK12:       omp.inner.for.inc:
10356 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10357 // CHECK12-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
10358 // CHECK12-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
10359 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
10360 // CHECK12:       omp.inner.for.end:
10361 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10362 // CHECK12:       omp.loop.exit:
10363 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
10364 // CHECK12-NEXT:    ret void
10365 //
10366 //
10367 // CHECK17-LABEL: define {{[^@]+}}@_Z3fooi
10368 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
10369 // CHECK17-NEXT:  entry:
10370 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10371 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
10372 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
10373 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
10374 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
10375 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
10376 // CHECK17-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
10377 // CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
10378 // CHECK17-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
10379 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10380 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
10381 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10382 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10383 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
10384 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
10385 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
10386 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
10387 // CHECK17-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
10388 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10389 // CHECK17-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
10390 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
10391 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
10392 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
10393 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10394 // CHECK17-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
10395 // CHECK17-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
10396 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
10397 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
10398 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
10399 // CHECK17-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
10400 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_22:%.*]] = alloca i32, align 4
10401 // CHECK17-NEXT:    [[A_CASTED23:%.*]] = alloca i64, align 8
10402 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED25:%.*]] = alloca i64, align 8
10403 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [10 x i8*], align 8
10404 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [10 x i8*], align 8
10405 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [10 x i8*], align 8
10406 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
10407 // CHECK17-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
10408 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
10409 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10410 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
10411 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
10412 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
10413 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
10414 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
10415 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
10416 // CHECK17-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
10417 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
10418 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
10419 // CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
10420 // CHECK17-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
10421 // CHECK17-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
10422 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
10423 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
10424 // CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
10425 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
10426 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
10427 // CHECK17-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
10428 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10429 // CHECK17-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
10430 // CHECK17-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10431 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10432 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
10433 // CHECK17-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
10434 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
10435 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10436 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
10437 // CHECK17-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
10438 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
10439 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10440 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
10441 // CHECK17-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
10442 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10443 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
10444 // CHECK17-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
10445 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10446 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
10447 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10448 // CHECK17-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
10449 // CHECK17-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
10450 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10451 // CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
10452 // CHECK17-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
10453 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
10454 // CHECK17-NEXT:    store i8* null, i8** [[TMP24]], align 8
10455 // CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10456 // CHECK17-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
10457 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
10458 // CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10459 // CHECK17-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
10460 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
10461 // CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
10462 // CHECK17-NEXT:    store i8* null, i8** [[TMP29]], align 8
10463 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10464 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10465 // CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
10466 // CHECK17-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
10467 // CHECK17-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
10468 // CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
10469 // CHECK17-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10470 // CHECK17-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
10471 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
10472 // CHECK17-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10473 // CHECK17-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
10474 // CHECK17-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 120, i64 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
10475 // CHECK17-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
10476 // CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
10477 // CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
10478 // CHECK17-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
10479 // CHECK17-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
10480 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
10481 // CHECK17-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
10482 // CHECK17-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
10483 // CHECK17-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
10484 // CHECK17-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
10485 // CHECK17-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
10486 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
10487 // CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
10488 // CHECK17-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
10489 // CHECK17-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
10490 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
10491 // CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
10492 // CHECK17-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
10493 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP53]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
10494 // CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
10495 // CHECK17-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
10496 // CHECK17-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
10497 // CHECK17-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP38]])
10498 // CHECK17-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
10499 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10500 // CHECK17-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
10501 // CHECK17-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
10502 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
10503 // CHECK17-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
10504 // CHECK17-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
10505 // CHECK17-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
10506 // CHECK17-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
10507 // CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
10508 // CHECK17-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
10509 // CHECK17-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
10510 // CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
10511 // CHECK17-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
10512 // CHECK17-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
10513 // CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
10514 // CHECK17-NEXT:    store i8* null, i8** [[TMP65]], align 8
10515 // CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
10516 // CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
10517 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
10518 // CHECK17-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10519 // CHECK17-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
10520 // CHECK17-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10521 // CHECK17:       omp_offload.failed:
10522 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i64 [[TMP60]]) #[[ATTR3]]
10523 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10524 // CHECK17:       omp_offload.cont:
10525 // CHECK17-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
10526 // CHECK17-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
10527 // CHECK17-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
10528 // CHECK17-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
10529 // CHECK17-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
10530 // CHECK17-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
10531 // CHECK17-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
10532 // CHECK17-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
10533 // CHECK17-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
10534 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
10535 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10536 // CHECK17:       omp_if.then:
10537 // CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
10538 // CHECK17-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
10539 // CHECK17-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
10540 // CHECK17-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
10541 // CHECK17-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
10542 // CHECK17-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
10543 // CHECK17-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
10544 // CHECK17-NEXT:    store i8* null, i8** [[TMP79]], align 8
10545 // CHECK17-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
10546 // CHECK17-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
10547 // CHECK17-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
10548 // CHECK17-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
10549 // CHECK17-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
10550 // CHECK17-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
10551 // CHECK17-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
10552 // CHECK17-NEXT:    store i8* null, i8** [[TMP84]], align 8
10553 // CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
10554 // CHECK17-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
10555 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
10556 // CHECK17-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP85]], i8** [[TMP86]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10557 // CHECK17-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
10558 // CHECK17-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
10559 // CHECK17:       omp_offload.failed20:
10560 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
10561 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
10562 // CHECK17:       omp_offload.cont21:
10563 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
10564 // CHECK17:       omp_if.else:
10565 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
10566 // CHECK17-NEXT:    br label [[OMP_IF_END]]
10567 // CHECK17:       omp_if.end:
10568 // CHECK17-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
10569 // CHECK17-NEXT:    store i32 [[TMP89]], i32* [[DOTCAPTURE_EXPR_22]], align 4
10570 // CHECK17-NEXT:    [[TMP90:%.*]] = load i32, i32* [[A]], align 4
10571 // CHECK17-NEXT:    [[CONV24:%.*]] = bitcast i64* [[A_CASTED23]] to i32*
10572 // CHECK17-NEXT:    store i32 [[TMP90]], i32* [[CONV24]], align 4
10573 // CHECK17-NEXT:    [[TMP91:%.*]] = load i64, i64* [[A_CASTED23]], align 8
10574 // CHECK17-NEXT:    [[TMP92:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_22]], align 4
10575 // CHECK17-NEXT:    [[CONV26:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED25]] to i32*
10576 // CHECK17-NEXT:    store i32 [[TMP92]], i32* [[CONV26]], align 4
10577 // CHECK17-NEXT:    [[TMP93:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED25]], align 8
10578 // CHECK17-NEXT:    [[TMP94:%.*]] = load i32, i32* [[N_ADDR]], align 4
10579 // CHECK17-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[TMP94]], 20
10580 // CHECK17-NEXT:    br i1 [[CMP27]], label [[OMP_IF_THEN28:%.*]], label [[OMP_IF_ELSE35:%.*]]
10581 // CHECK17:       omp_if.then28:
10582 // CHECK17-NEXT:    [[TMP95:%.*]] = mul nuw i64 [[TMP2]], 4
10583 // CHECK17-NEXT:    [[TMP96:%.*]] = mul nuw i64 5, [[TMP5]]
10584 // CHECK17-NEXT:    [[TMP97:%.*]] = mul nuw i64 [[TMP96]], 8
10585 // CHECK17-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
10586 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP98]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i64 80, i1 false)
10587 // CHECK17-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
10588 // CHECK17-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i64*
10589 // CHECK17-NEXT:    store i64 [[TMP91]], i64* [[TMP100]], align 8
10590 // CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
10591 // CHECK17-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i64*
10592 // CHECK17-NEXT:    store i64 [[TMP91]], i64* [[TMP102]], align 8
10593 // CHECK17-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
10594 // CHECK17-NEXT:    store i8* null, i8** [[TMP103]], align 8
10595 // CHECK17-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
10596 // CHECK17-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
10597 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 8
10598 // CHECK17-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
10599 // CHECK17-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
10600 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 8
10601 // CHECK17-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 1
10602 // CHECK17-NEXT:    store i8* null, i8** [[TMP108]], align 8
10603 // CHECK17-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 2
10604 // CHECK17-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i64*
10605 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP110]], align 8
10606 // CHECK17-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 2
10607 // CHECK17-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i64*
10608 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP112]], align 8
10609 // CHECK17-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 2
10610 // CHECK17-NEXT:    store i8* null, i8** [[TMP113]], align 8
10611 // CHECK17-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 3
10612 // CHECK17-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
10613 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP115]], align 8
10614 // CHECK17-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 3
10615 // CHECK17-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
10616 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP117]], align 8
10617 // CHECK17-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
10618 // CHECK17-NEXT:    store i64 [[TMP95]], i64* [[TMP118]], align 8
10619 // CHECK17-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 3
10620 // CHECK17-NEXT:    store i8* null, i8** [[TMP119]], align 8
10621 // CHECK17-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 4
10622 // CHECK17-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
10623 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 8
10624 // CHECK17-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 4
10625 // CHECK17-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
10626 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 8
10627 // CHECK17-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 4
10628 // CHECK17-NEXT:    store i8* null, i8** [[TMP124]], align 8
10629 // CHECK17-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 5
10630 // CHECK17-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i64*
10631 // CHECK17-NEXT:    store i64 5, i64* [[TMP126]], align 8
10632 // CHECK17-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 5
10633 // CHECK17-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
10634 // CHECK17-NEXT:    store i64 5, i64* [[TMP128]], align 8
10635 // CHECK17-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 5
10636 // CHECK17-NEXT:    store i8* null, i8** [[TMP129]], align 8
10637 // CHECK17-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 6
10638 // CHECK17-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
10639 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP131]], align 8
10640 // CHECK17-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 6
10641 // CHECK17-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i64*
10642 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP133]], align 8
10643 // CHECK17-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 6
10644 // CHECK17-NEXT:    store i8* null, i8** [[TMP134]], align 8
10645 // CHECK17-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 7
10646 // CHECK17-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
10647 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 8
10648 // CHECK17-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 7
10649 // CHECK17-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
10650 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 8
10651 // CHECK17-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
10652 // CHECK17-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 8
10653 // CHECK17-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 7
10654 // CHECK17-NEXT:    store i8* null, i8** [[TMP140]], align 8
10655 // CHECK17-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 8
10656 // CHECK17-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
10657 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 8
10658 // CHECK17-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 8
10659 // CHECK17-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
10660 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 8
10661 // CHECK17-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 8
10662 // CHECK17-NEXT:    store i8* null, i8** [[TMP145]], align 8
10663 // CHECK17-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 9
10664 // CHECK17-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i64*
10665 // CHECK17-NEXT:    store i64 [[TMP93]], i64* [[TMP147]], align 8
10666 // CHECK17-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 9
10667 // CHECK17-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
10668 // CHECK17-NEXT:    store i64 [[TMP93]], i64* [[TMP149]], align 8
10669 // CHECK17-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 9
10670 // CHECK17-NEXT:    store i8* null, i8** [[TMP150]], align 8
10671 // CHECK17-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
10672 // CHECK17-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
10673 // CHECK17-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
10674 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
10675 // CHECK17-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10676 // CHECK17-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
10677 // CHECK17-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED33:%.*]], label [[OMP_OFFLOAD_CONT34:%.*]]
10678 // CHECK17:       omp_offload.failed33:
10679 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
10680 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT34]]
10681 // CHECK17:       omp_offload.cont34:
10682 // CHECK17-NEXT:    br label [[OMP_IF_END36:%.*]]
10683 // CHECK17:       omp_if.else35:
10684 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
10685 // CHECK17-NEXT:    br label [[OMP_IF_END36]]
10686 // CHECK17:       omp_if.end36:
10687 // CHECK17-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
10688 // CHECK17-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
10689 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
10690 // CHECK17-NEXT:    ret i32 [[TMP156]]
10691 //
10692 //
10693 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
10694 // CHECK17-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
10695 // CHECK17-NEXT:  entry:
10696 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10697 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10698 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
10699 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10700 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
10701 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10702 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10703 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
10704 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10705 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
10706 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
10707 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
10708 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
10709 // CHECK17-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
10710 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
10711 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10712 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
10713 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10714 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
10715 // CHECK17-NEXT:    ret void
10716 //
10717 //
10718 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
10719 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
10720 // CHECK17-NEXT:  entry:
10721 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10722 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10723 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10724 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10725 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10726 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10727 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10728 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10729 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10730 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
10731 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10732 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10733 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10734 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10735 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10736 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
10737 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10738 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10739 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10740 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10741 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10742 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10743 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
10744 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10745 // CHECK17:       cond.true:
10746 // CHECK17-NEXT:    br label [[COND_END:%.*]]
10747 // CHECK17:       cond.false:
10748 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10749 // CHECK17-NEXT:    br label [[COND_END]]
10750 // CHECK17:       cond.end:
10751 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10752 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10753 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10754 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10755 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10756 // CHECK17:       omp.inner.for.cond:
10757 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10758 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10759 // CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10760 // CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10761 // CHECK17:       omp.inner.for.body:
10762 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10763 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
10764 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10765 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
10766 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10767 // CHECK17:       omp.body.continue:
10768 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10769 // CHECK17:       omp.inner.for.inc:
10770 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10771 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
10772 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
10773 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
10774 // CHECK17:       omp.inner.for.end:
10775 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10776 // CHECK17:       omp.loop.exit:
10777 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10778 // CHECK17-NEXT:    ret void
10779 //
10780 //
10781 // CHECK17-LABEL: define {{[^@]+}}@.omp_task_privates_map.
10782 // CHECK17-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
10783 // CHECK17-NEXT:  entry:
10784 // CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
10785 // CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
10786 // CHECK17-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
10787 // CHECK17-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
10788 // CHECK17-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
10789 // CHECK17-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
10790 // CHECK17-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
10791 // CHECK17-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
10792 // CHECK17-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
10793 // CHECK17-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
10794 // CHECK17-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
10795 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
10796 // CHECK17-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
10797 // CHECK17-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
10798 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
10799 // CHECK17-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
10800 // CHECK17-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
10801 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
10802 // CHECK17-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
10803 // CHECK17-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
10804 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
10805 // CHECK17-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
10806 // CHECK17-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
10807 // CHECK17-NEXT:    ret void
10808 //
10809 //
10810 // CHECK17-LABEL: define {{[^@]+}}@.omp_task_entry.
10811 // CHECK17-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
10812 // CHECK17-NEXT:  entry:
10813 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
10814 // CHECK17-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
10815 // CHECK17-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
10816 // CHECK17-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
10817 // CHECK17-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
10818 // CHECK17-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
10819 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
10820 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
10821 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
10822 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
10823 // CHECK17-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
10824 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
10825 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
10826 // CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
10827 // CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
10828 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
10829 // CHECK17-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
10830 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
10831 // CHECK17-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
10832 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
10833 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
10834 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
10835 // CHECK17-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
10836 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
10837 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
10838 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
10839 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
10840 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
10841 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
10842 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
10843 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
10844 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
10845 // CHECK17-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
10846 // CHECK17-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
10847 // CHECK17-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
10848 // CHECK17-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
10849 // CHECK17-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
10850 // CHECK17-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
10851 // CHECK17-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
10852 // CHECK17-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
10853 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
10854 // CHECK17-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
10855 // CHECK17-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !21
10856 // CHECK17-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !21
10857 // CHECK17-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !21
10858 // CHECK17-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !21
10859 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
10860 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
10861 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
10862 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
10863 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
10864 // CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
10865 // CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
10866 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
10867 // CHECK17-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
10868 // CHECK17-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
10869 // CHECK17-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
10870 // CHECK17:       omp_offload.failed.i:
10871 // CHECK17-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
10872 // CHECK17-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
10873 // CHECK17-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !21
10874 // CHECK17-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !21
10875 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
10876 // CHECK17-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
10877 // CHECK17-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !21
10878 // CHECK17-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !21
10879 // CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
10880 // CHECK17-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
10881 // CHECK17-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !21
10882 // CHECK17-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !21
10883 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
10884 // CHECK17-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
10885 // CHECK17:       .omp_outlined..1.exit:
10886 // CHECK17-NEXT:    ret i32 0
10887 //
10888 //
10889 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
10890 // CHECK17-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
10891 // CHECK17-NEXT:  entry:
10892 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10893 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10894 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10895 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10896 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
10897 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10898 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
10899 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
10900 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
10901 // CHECK17-NEXT:    ret void
10902 //
10903 //
10904 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
10905 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
10906 // CHECK17-NEXT:  entry:
10907 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10908 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10909 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10910 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10911 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10912 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10913 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10914 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10915 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10916 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
10917 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10918 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10919 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10920 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10921 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10922 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
10923 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10924 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10925 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10926 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10927 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10928 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10929 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
10930 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10931 // CHECK17:       cond.true:
10932 // CHECK17-NEXT:    br label [[COND_END:%.*]]
10933 // CHECK17:       cond.false:
10934 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10935 // CHECK17-NEXT:    br label [[COND_END]]
10936 // CHECK17:       cond.end:
10937 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10938 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10939 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10940 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10941 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10942 // CHECK17:       omp.inner.for.cond:
10943 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10944 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10945 // CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10946 // CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10947 // CHECK17:       omp.inner.for.body:
10948 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10949 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
10950 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10951 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
10952 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
10953 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
10954 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[CONV]], align 4
10955 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10956 // CHECK17:       omp.body.continue:
10957 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10958 // CHECK17:       omp.inner.for.inc:
10959 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10960 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
10961 // CHECK17-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
10962 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
10963 // CHECK17:       omp.inner.for.end:
10964 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10965 // CHECK17:       omp.loop.exit:
10966 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10967 // CHECK17-NEXT:    ret void
10968 //
10969 //
10970 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
10971 // CHECK17-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
10972 // CHECK17-NEXT:  entry:
10973 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10974 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10975 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10976 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10977 // CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
10978 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10979 // CHECK17-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
10980 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10981 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
10982 // CHECK17-NEXT:    ret void
10983 //
10984 //
10985 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
10986 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
10987 // CHECK17-NEXT:  entry:
10988 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10989 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10990 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10991 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10992 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10993 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10994 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10995 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10996 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10997 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
10998 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10999 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11000 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11001 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11002 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11003 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
11004 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11005 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11006 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11007 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11008 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11009 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11010 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
11011 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11012 // CHECK17:       cond.true:
11013 // CHECK17-NEXT:    br label [[COND_END:%.*]]
11014 // CHECK17:       cond.false:
11015 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11016 // CHECK17-NEXT:    br label [[COND_END]]
11017 // CHECK17:       cond.end:
11018 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11019 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11020 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11021 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11022 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11023 // CHECK17:       omp.inner.for.cond:
11024 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11025 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11026 // CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11027 // CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11028 // CHECK17:       omp.inner.for.body:
11029 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11030 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
11031 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11032 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
11033 // CHECK17-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
11034 // CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
11035 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
11036 // CHECK17-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
11037 // CHECK17-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
11038 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11039 // CHECK17:       omp.body.continue:
11040 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11041 // CHECK17:       omp.inner.for.inc:
11042 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11043 // CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
11044 // CHECK17-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
11045 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
11046 // CHECK17:       omp.inner.for.end:
11047 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11048 // CHECK17:       omp.loop.exit:
11049 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11050 // CHECK17-NEXT:    ret void
11051 //
11052 //
11053 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
11054 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
11055 // CHECK17-NEXT:  entry:
11056 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11057 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11058 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11059 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11060 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11061 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11062 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11063 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11064 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
11065 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11066 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
11067 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
11068 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
11069 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11070 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
11071 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11072 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
11073 // CHECK17-NEXT:    ret void
11074 //
11075 //
11076 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
11077 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
11078 // CHECK17-NEXT:  entry:
11079 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11080 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11081 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11082 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11083 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11084 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11085 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11086 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11087 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11088 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11089 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
11090 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11091 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11092 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11093 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11094 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11095 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11096 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11097 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
11098 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11099 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11100 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11101 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11102 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11103 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11104 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
11105 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11106 // CHECK17:       cond.true:
11107 // CHECK17-NEXT:    br label [[COND_END:%.*]]
11108 // CHECK17:       cond.false:
11109 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11110 // CHECK17-NEXT:    br label [[COND_END]]
11111 // CHECK17:       cond.end:
11112 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11113 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11114 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11115 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11116 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11117 // CHECK17:       omp.inner.for.cond:
11118 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11119 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11120 // CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11121 // CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11122 // CHECK17:       omp.inner.for.body:
11123 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11124 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
11125 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11126 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
11127 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
11128 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
11129 // CHECK17-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
11130 // CHECK17-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
11131 // CHECK17-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
11132 // CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
11133 // CHECK17-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
11134 // CHECK17-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
11135 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11136 // CHECK17:       omp.body.continue:
11137 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11138 // CHECK17:       omp.inner.for.inc:
11139 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11140 // CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
11141 // CHECK17-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
11142 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
11143 // CHECK17:       omp.inner.for.end:
11144 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11145 // CHECK17:       omp.loop.exit:
11146 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11147 // CHECK17-NEXT:    ret void
11148 //
11149 //
11150 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
11151 // CHECK17-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11152 // CHECK17-NEXT:  entry:
11153 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11154 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
11155 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11156 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
11157 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
11158 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11159 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
11160 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
11161 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
11162 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11163 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11164 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
11165 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11166 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
11167 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11168 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
11169 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
11170 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11171 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
11172 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
11173 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
11174 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11175 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11176 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
11177 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11178 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
11179 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
11180 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11181 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
11182 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
11183 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
11184 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11185 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
11186 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11187 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
11188 // CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
11189 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
11190 // CHECK17-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
11191 // CHECK17-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
11192 // CHECK17-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
11193 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
11194 // CHECK17-NEXT:    ret void
11195 //
11196 //
11197 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
11198 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11199 // CHECK17-NEXT:  entry:
11200 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11201 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11202 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11203 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
11204 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11205 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
11206 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
11207 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11208 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
11209 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
11210 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
11211 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11212 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11213 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11214 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11215 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11216 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11217 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11218 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
11219 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11220 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11221 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11222 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
11223 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11224 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
11225 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
11226 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11227 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
11228 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
11229 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
11230 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11231 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11232 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
11233 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11234 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
11235 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
11236 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11237 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
11238 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
11239 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
11240 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11241 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11242 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
11243 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11244 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11245 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
11246 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11247 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
11248 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
11249 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
11250 // CHECK17:       omp.dispatch.cond:
11251 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11252 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
11253 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11254 // CHECK17:       cond.true:
11255 // CHECK17-NEXT:    br label [[COND_END:%.*]]
11256 // CHECK17:       cond.false:
11257 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11258 // CHECK17-NEXT:    br label [[COND_END]]
11259 // CHECK17:       cond.end:
11260 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
11261 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11262 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11263 // CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
11264 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11265 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11266 // CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
11267 // CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
11268 // CHECK17:       omp.dispatch.body:
11269 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11270 // CHECK17:       omp.inner.for.cond:
11271 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
11272 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
11273 // CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
11274 // CHECK17-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11275 // CHECK17:       omp.inner.for.body:
11276 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
11277 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
11278 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11279 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
11280 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !22
11281 // CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
11282 // CHECK17-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !22
11283 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
11284 // CHECK17-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !22
11285 // CHECK17-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
11286 // CHECK17-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
11287 // CHECK17-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
11288 // CHECK17-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !22
11289 // CHECK17-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
11290 // CHECK17-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
11291 // CHECK17-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
11292 // CHECK17-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
11293 // CHECK17-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
11294 // CHECK17-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
11295 // CHECK17-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
11296 // CHECK17-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
11297 // CHECK17-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
11298 // CHECK17-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
11299 // CHECK17-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
11300 // CHECK17-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
11301 // CHECK17-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
11302 // CHECK17-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
11303 // CHECK17-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
11304 // CHECK17-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
11305 // CHECK17-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
11306 // CHECK17-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
11307 // CHECK17-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !22
11308 // CHECK17-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
11309 // CHECK17-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !22
11310 // CHECK17-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
11311 // CHECK17-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !22
11312 // CHECK17-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
11313 // CHECK17-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
11314 // CHECK17-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
11315 // CHECK17-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !22
11316 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11317 // CHECK17:       omp.body.continue:
11318 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11319 // CHECK17:       omp.inner.for.inc:
11320 // CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
11321 // CHECK17-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
11322 // CHECK17-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
11323 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
11324 // CHECK17:       omp.inner.for.end:
11325 // CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
11326 // CHECK17:       omp.dispatch.inc:
11327 // CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11328 // CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11329 // CHECK17-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
11330 // CHECK17-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
11331 // CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11332 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11333 // CHECK17-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
11334 // CHECK17-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
11335 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
11336 // CHECK17:       omp.dispatch.end:
11337 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
11338 // CHECK17-NEXT:    ret void
11339 //
11340 //
11341 // CHECK17-LABEL: define {{[^@]+}}@_Z3bari
11342 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
11343 // CHECK17-NEXT:  entry:
11344 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11345 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
11346 // CHECK17-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
11347 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11348 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
11349 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11350 // CHECK17-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
11351 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
11352 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
11353 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11354 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
11355 // CHECK17-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
11356 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
11357 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
11358 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
11359 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
11360 // CHECK17-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
11361 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11362 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
11363 // CHECK17-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
11364 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
11365 // CHECK17-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
11366 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
11367 // CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
11368 // CHECK17-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
11369 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
11370 // CHECK17-NEXT:    ret i32 [[TMP8]]
11371 //
11372 //
11373 // CHECK17-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
11374 // CHECK17-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
11375 // CHECK17-NEXT:  entry:
11376 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
11377 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11378 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
11379 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
11380 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
11381 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
11382 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
11383 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
11384 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
11385 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
11386 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11387 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
11388 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11389 // CHECK17-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
11390 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11391 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
11392 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
11393 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
11394 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
11395 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
11396 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
11397 // CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
11398 // CHECK17-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
11399 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
11400 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
11401 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
11402 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
11403 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
11404 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
11405 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
11406 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11407 // CHECK17:       omp_if.then:
11408 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
11409 // CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
11410 // CHECK17-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
11411 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
11412 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i64 40, i1 false)
11413 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11414 // CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
11415 // CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
11416 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11417 // CHECK17-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
11418 // CHECK17-NEXT:    store double* [[A]], double** [[TMP14]], align 8
11419 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
11420 // CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
11421 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11422 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
11423 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
11424 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11425 // CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
11426 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
11427 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
11428 // CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
11429 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11430 // CHECK17-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
11431 // CHECK17-NEXT:    store i64 2, i64* [[TMP22]], align 8
11432 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11433 // CHECK17-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
11434 // CHECK17-NEXT:    store i64 2, i64* [[TMP24]], align 8
11435 // CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
11436 // CHECK17-NEXT:    store i8* null, i8** [[TMP25]], align 8
11437 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
11438 // CHECK17-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
11439 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
11440 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
11441 // CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
11442 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
11443 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
11444 // CHECK17-NEXT:    store i8* null, i8** [[TMP30]], align 8
11445 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
11446 // CHECK17-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
11447 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
11448 // CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
11449 // CHECK17-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
11450 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
11451 // CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
11452 // CHECK17-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
11453 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
11454 // CHECK17-NEXT:    store i8* null, i8** [[TMP36]], align 8
11455 // CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11456 // CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11457 // CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
11458 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
11459 // CHECK17-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11460 // CHECK17-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
11461 // CHECK17-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11462 // CHECK17:       omp_offload.failed:
11463 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
11464 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11465 // CHECK17:       omp_offload.cont:
11466 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
11467 // CHECK17:       omp_if.else:
11468 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
11469 // CHECK17-NEXT:    br label [[OMP_IF_END]]
11470 // CHECK17:       omp_if.end:
11471 // CHECK17-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
11472 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
11473 // CHECK17-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
11474 // CHECK17-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
11475 // CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
11476 // CHECK17-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
11477 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
11478 // CHECK17-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
11479 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
11480 // CHECK17-NEXT:    ret i32 [[ADD4]]
11481 //
11482 //
11483 // CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
11484 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
11485 // CHECK17-NEXT:  entry:
11486 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11487 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
11488 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
11489 // CHECK17-NEXT:    [[AAA:%.*]] = alloca i8, align 1
11490 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
11491 // CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
11492 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11493 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11494 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
11495 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
11496 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
11497 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
11498 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11499 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11500 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
11501 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
11502 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11503 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
11504 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
11505 // CHECK17-NEXT:    store i8 0, i8* [[AAA]], align 1
11506 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11507 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
11508 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
11509 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[N_CASTED]], align 8
11510 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
11511 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11512 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
11513 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
11514 // CHECK17-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
11515 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11516 // CHECK17-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
11517 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11518 // CHECK17-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
11519 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
11520 // CHECK17-NEXT:    store i8 [[TMP6]], i8* [[CONV3]], align 1
11521 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
11522 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
11523 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
11524 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11525 // CHECK17:       omp_if.then:
11526 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11527 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
11528 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
11529 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11530 // CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
11531 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP12]], align 8
11532 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
11533 // CHECK17-NEXT:    store i8* null, i8** [[TMP13]], align 8
11534 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11535 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
11536 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
11537 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11538 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
11539 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP17]], align 8
11540 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
11541 // CHECK17-NEXT:    store i8* null, i8** [[TMP18]], align 8
11542 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11543 // CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
11544 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
11545 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11546 // CHECK17-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
11547 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
11548 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
11549 // CHECK17-NEXT:    store i8* null, i8** [[TMP23]], align 8
11550 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
11551 // CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
11552 // CHECK17-NEXT:    store i64 [[TMP7]], i64* [[TMP25]], align 8
11553 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
11554 // CHECK17-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
11555 // CHECK17-NEXT:    store i64 [[TMP7]], i64* [[TMP27]], align 8
11556 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
11557 // CHECK17-NEXT:    store i8* null, i8** [[TMP28]], align 8
11558 // CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
11559 // CHECK17-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
11560 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 8
11561 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
11562 // CHECK17-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
11563 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 8
11564 // CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
11565 // CHECK17-NEXT:    store i8* null, i8** [[TMP33]], align 8
11566 // CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11567 // CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11568 // CHECK17-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
11569 // CHECK17-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
11570 // CHECK17-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
11571 // CHECK17-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_4]], align 4
11572 // CHECK17-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
11573 // CHECK17-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11574 // CHECK17-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
11575 // CHECK17-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
11576 // CHECK17-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
11577 // CHECK17-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
11578 // CHECK17-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
11579 // CHECK17-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
11580 // CHECK17-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
11581 // CHECK17-NEXT:    [[ADD8:%.*]] = add i32 [[TMP40]], 1
11582 // CHECK17-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD8]] to i64
11583 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
11584 // CHECK17-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11585 // CHECK17-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
11586 // CHECK17-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11587 // CHECK17:       omp_offload.failed:
11588 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
11589 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11590 // CHECK17:       omp_offload.cont:
11591 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
11592 // CHECK17:       omp_if.else:
11593 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
11594 // CHECK17-NEXT:    br label [[OMP_IF_END]]
11595 // CHECK17:       omp_if.end:
11596 // CHECK17-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
11597 // CHECK17-NEXT:    ret i32 [[TMP44]]
11598 //
11599 //
11600 // CHECK17-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
11601 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
11602 // CHECK17-NEXT:  entry:
11603 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11604 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
11605 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
11606 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
11607 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11608 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11609 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
11610 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
11611 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
11612 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11613 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11614 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
11615 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
11616 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
11617 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11618 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
11619 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
11620 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
11621 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11622 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
11623 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11624 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
11625 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
11626 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11627 // CHECK17:       omp_if.then:
11628 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11629 // CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
11630 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
11631 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11632 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
11633 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
11634 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
11635 // CHECK17-NEXT:    store i8* null, i8** [[TMP9]], align 8
11636 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11637 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
11638 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
11639 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11640 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
11641 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
11642 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
11643 // CHECK17-NEXT:    store i8* null, i8** [[TMP14]], align 8
11644 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11645 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
11646 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
11647 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11648 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
11649 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
11650 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
11651 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
11652 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11653 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11654 // CHECK17-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
11655 // CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11656 // CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
11657 // CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11658 // CHECK17:       omp_offload.failed:
11659 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
11660 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11661 // CHECK17:       omp_offload.cont:
11662 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
11663 // CHECK17:       omp_if.else:
11664 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
11665 // CHECK17-NEXT:    br label [[OMP_IF_END]]
11666 // CHECK17:       omp_if.end:
11667 // CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
11668 // CHECK17-NEXT:    ret i32 [[TMP24]]
11669 //
11670 //
11671 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
11672 // CHECK17-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
11673 // CHECK17-NEXT:  entry:
11674 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
11675 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
11676 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11677 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11678 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
11679 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
11680 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
11681 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
11682 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11683 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11684 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
11685 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
11686 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
11687 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11688 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11689 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
11690 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
11691 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
11692 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
11693 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
11694 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
11695 // CHECK17-NEXT:    ret void
11696 //
11697 //
11698 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..12
11699 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
11700 // CHECK17-NEXT:  entry:
11701 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11702 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11703 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
11704 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
11705 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
11706 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
11707 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
11708 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11709 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11710 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11711 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11712 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11713 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11714 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
11715 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11716 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11717 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
11718 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
11719 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11720 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
11721 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
11722 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
11723 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
11724 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11725 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
11726 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
11727 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11728 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
11729 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11730 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11731 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11732 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
11733 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11734 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11735 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
11736 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11737 // CHECK17:       cond.true:
11738 // CHECK17-NEXT:    br label [[COND_END:%.*]]
11739 // CHECK17:       cond.false:
11740 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11741 // CHECK17-NEXT:    br label [[COND_END]]
11742 // CHECK17:       cond.end:
11743 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
11744 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11745 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11746 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
11747 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11748 // CHECK17:       omp.inner.for.cond:
11749 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11750 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11751 // CHECK17-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
11752 // CHECK17-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11753 // CHECK17:       omp.inner.for.body:
11754 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11755 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
11756 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11757 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
11758 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
11759 // CHECK17-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
11760 // CHECK17-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
11761 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
11762 // CHECK17-NEXT:    store double [[ADD5]], double* [[A]], align 8
11763 // CHECK17-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11764 // CHECK17-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
11765 // CHECK17-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
11766 // CHECK17-NEXT:    store double [[INC]], double* [[A6]], align 8
11767 // CHECK17-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
11768 // CHECK17-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
11769 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
11770 // CHECK17-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
11771 // CHECK17-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
11772 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11773 // CHECK17:       omp.body.continue:
11774 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11775 // CHECK17:       omp.inner.for.inc:
11776 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11777 // CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
11778 // CHECK17-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
11779 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
11780 // CHECK17:       omp.inner.for.end:
11781 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11782 // CHECK17:       omp.loop.exit:
11783 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
11784 // CHECK17-NEXT:    ret void
11785 //
11786 //
11787 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
11788 // CHECK17-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11789 // CHECK17-NEXT:  entry:
11790 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
11791 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11792 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11793 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
11794 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11795 // CHECK17-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
11796 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11797 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11798 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
11799 // CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
11800 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11801 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11802 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
11803 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11804 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
11805 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11806 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11807 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
11808 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11809 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
11810 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
11811 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
11812 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
11813 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
11814 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11815 // CHECK17-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
11816 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
11817 // CHECK17-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
11818 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11819 // CHECK17-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
11820 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11821 // CHECK17-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
11822 // CHECK17-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
11823 // CHECK17-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
11824 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
11825 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
11826 // CHECK17-NEXT:    ret void
11827 //
11828 //
11829 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..15
11830 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11831 // CHECK17-NEXT:  entry:
11832 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11833 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11834 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
11835 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11836 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11837 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
11838 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11839 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11840 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11841 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11842 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
11843 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
11844 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
11845 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11846 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11847 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11848 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11849 // CHECK17-NEXT:    [[I8:%.*]] = alloca i32, align 4
11850 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11851 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11852 // CHECK17-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
11853 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11854 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11855 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
11856 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11857 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
11858 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11859 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11860 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
11861 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11862 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
11863 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
11864 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
11865 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
11866 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
11867 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11868 // CHECK17-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
11869 // CHECK17-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
11870 // CHECK17-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
11871 // CHECK17-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
11872 // CHECK17-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
11873 // CHECK17-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
11874 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11875 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
11876 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11877 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
11878 // CHECK17-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
11879 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
11880 // CHECK17:       omp.precond.then:
11881 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11882 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
11883 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
11884 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11885 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11886 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11887 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
11888 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11889 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11890 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
11891 // CHECK17-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
11892 // CHECK17-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11893 // CHECK17:       cond.true:
11894 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
11895 // CHECK17-NEXT:    br label [[COND_END:%.*]]
11896 // CHECK17:       cond.false:
11897 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11898 // CHECK17-NEXT:    br label [[COND_END]]
11899 // CHECK17:       cond.end:
11900 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
11901 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11902 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11903 // CHECK17-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
11904 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11905 // CHECK17:       omp.inner.for.cond:
11906 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11907 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11908 // CHECK17-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
11909 // CHECK17-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
11910 // CHECK17-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11911 // CHECK17:       omp.inner.for.body:
11912 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11913 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11914 // CHECK17-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
11915 // CHECK17-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
11916 // CHECK17-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
11917 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
11918 // CHECK17-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
11919 // CHECK17-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
11920 // CHECK17-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
11921 // CHECK17-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
11922 // CHECK17-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
11923 // CHECK17-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
11924 // CHECK17-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
11925 // CHECK17-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
11926 // CHECK17-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
11927 // CHECK17-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
11928 // CHECK17-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
11929 // CHECK17-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
11930 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
11931 // CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11932 // CHECK17-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
11933 // CHECK17-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
11934 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11935 // CHECK17:       omp.body.continue:
11936 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11937 // CHECK17:       omp.inner.for.inc:
11938 // CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11939 // CHECK17-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
11940 // CHECK17-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
11941 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
11942 // CHECK17:       omp.inner.for.end:
11943 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11944 // CHECK17:       omp.loop.exit:
11945 // CHECK17-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11946 // CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
11947 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
11948 // CHECK17-NEXT:    br label [[OMP_PRECOND_END]]
11949 // CHECK17:       omp.precond.end:
11950 // CHECK17-NEXT:    ret void
11951 //
11952 //
11953 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
11954 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11955 // CHECK17-NEXT:  entry:
11956 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11957 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11958 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11959 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11960 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11961 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11962 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11963 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11964 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11965 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11966 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11967 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
11968 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11969 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
11970 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
11971 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
11972 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11973 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
11974 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11975 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
11976 // CHECK17-NEXT:    ret void
11977 //
11978 //
11979 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..18
11980 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11981 // CHECK17-NEXT:  entry:
11982 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11983 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11984 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11985 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11986 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11987 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11988 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11989 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11990 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11991 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11992 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11993 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
11994 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11995 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11996 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11997 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11998 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11999 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12000 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12001 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
12002 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12003 // CHECK17-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12004 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12005 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12006 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12007 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12008 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12009 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12010 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
12011 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12012 // CHECK17:       cond.true:
12013 // CHECK17-NEXT:    br label [[COND_END:%.*]]
12014 // CHECK17:       cond.false:
12015 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12016 // CHECK17-NEXT:    br label [[COND_END]]
12017 // CHECK17:       cond.end:
12018 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
12019 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12020 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12021 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
12022 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12023 // CHECK17:       omp.inner.for.cond:
12024 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12025 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12026 // CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
12027 // CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12028 // CHECK17:       omp.inner.for.body:
12029 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12030 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
12031 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12032 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12033 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
12034 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
12035 // CHECK17-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
12036 // CHECK17-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
12037 // CHECK17-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
12038 // CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
12039 // CHECK17-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
12040 // CHECK17-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
12041 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
12042 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
12043 // CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
12044 // CHECK17-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
12045 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12046 // CHECK17:       omp.body.continue:
12047 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12048 // CHECK17:       omp.inner.for.inc:
12049 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12050 // CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
12051 // CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
12052 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]]
12053 // CHECK17:       omp.inner.for.end:
12054 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12055 // CHECK17:       omp.loop.exit:
12056 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
12057 // CHECK17-NEXT:    ret void
12058 //
12059 //
12060 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
12061 // CHECK17-SAME: () #[[ATTR4]] {
12062 // CHECK17-NEXT:  entry:
12063 // CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
12064 // CHECK17-NEXT:    ret void
12065 //
12066 //
12067 // CHECK18-LABEL: define {{[^@]+}}@_Z3fooi
12068 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
12069 // CHECK18-NEXT:  entry:
12070 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12071 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
12072 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
12073 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
12074 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
12075 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
12076 // CHECK18-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
12077 // CHECK18-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
12078 // CHECK18-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
12079 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12080 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
12081 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
12082 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
12083 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
12084 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
12085 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
12086 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
12087 // CHECK18-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
12088 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
12089 // CHECK18-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
12090 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
12091 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
12092 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
12093 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12094 // CHECK18-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
12095 // CHECK18-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
12096 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
12097 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
12098 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
12099 // CHECK18-NEXT:    [[_TMP19:%.*]] = alloca i32, align 4
12100 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_22:%.*]] = alloca i32, align 4
12101 // CHECK18-NEXT:    [[A_CASTED23:%.*]] = alloca i64, align 8
12102 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED25:%.*]] = alloca i64, align 8
12103 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS29:%.*]] = alloca [10 x i8*], align 8
12104 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS30:%.*]] = alloca [10 x i8*], align 8
12105 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS31:%.*]] = alloca [10 x i8*], align 8
12106 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
12107 // CHECK18-NEXT:    [[_TMP32:%.*]] = alloca i32, align 4
12108 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
12109 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12110 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
12111 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
12112 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
12113 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
12114 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
12115 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
12116 // CHECK18-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
12117 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
12118 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
12119 // CHECK18-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
12120 // CHECK18-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
12121 // CHECK18-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
12122 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
12123 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
12124 // CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
12125 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
12126 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
12127 // CHECK18-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
12128 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
12129 // CHECK18-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
12130 // CHECK18-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
12131 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12132 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
12133 // CHECK18-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
12134 // CHECK18-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
12135 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
12136 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
12137 // CHECK18-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
12138 // CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
12139 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12140 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
12141 // CHECK18-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
12142 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12143 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
12144 // CHECK18-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
12145 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
12146 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
12147 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
12148 // CHECK18-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
12149 // CHECK18-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
12150 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
12151 // CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
12152 // CHECK18-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
12153 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
12154 // CHECK18-NEXT:    store i8* null, i8** [[TMP24]], align 8
12155 // CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
12156 // CHECK18-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
12157 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
12158 // CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
12159 // CHECK18-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
12160 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
12161 // CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
12162 // CHECK18-NEXT:    store i8* null, i8** [[TMP29]], align 8
12163 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12164 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12165 // CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
12166 // CHECK18-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
12167 // CHECK18-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
12168 // CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
12169 // CHECK18-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12170 // CHECK18-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
12171 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
12172 // CHECK18-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
12173 // CHECK18-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
12174 // CHECK18-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 120, i64 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
12175 // CHECK18-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
12176 // CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
12177 // CHECK18-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
12178 // CHECK18-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
12179 // CHECK18-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
12180 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
12181 // CHECK18-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
12182 // CHECK18-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
12183 // CHECK18-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
12184 // CHECK18-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
12185 // CHECK18-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
12186 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
12187 // CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
12188 // CHECK18-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
12189 // CHECK18-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
12190 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
12191 // CHECK18-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
12192 // CHECK18-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
12193 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP53]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
12194 // CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
12195 // CHECK18-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
12196 // CHECK18-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
12197 // CHECK18-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP38]])
12198 // CHECK18-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
12199 // CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
12200 // CHECK18-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
12201 // CHECK18-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
12202 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
12203 // CHECK18-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
12204 // CHECK18-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
12205 // CHECK18-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
12206 // CHECK18-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
12207 // CHECK18-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
12208 // CHECK18-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
12209 // CHECK18-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
12210 // CHECK18-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
12211 // CHECK18-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
12212 // CHECK18-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
12213 // CHECK18-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
12214 // CHECK18-NEXT:    store i8* null, i8** [[TMP65]], align 8
12215 // CHECK18-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
12216 // CHECK18-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
12217 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
12218 // CHECK18-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP66]], i8** [[TMP67]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12219 // CHECK18-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
12220 // CHECK18-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12221 // CHECK18:       omp_offload.failed:
12222 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i64 [[TMP60]]) #[[ATTR3]]
12223 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12224 // CHECK18:       omp_offload.cont:
12225 // CHECK18-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
12226 // CHECK18-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
12227 // CHECK18-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
12228 // CHECK18-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
12229 // CHECK18-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
12230 // CHECK18-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
12231 // CHECK18-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
12232 // CHECK18-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
12233 // CHECK18-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
12234 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
12235 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12236 // CHECK18:       omp_if.then:
12237 // CHECK18-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
12238 // CHECK18-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
12239 // CHECK18-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
12240 // CHECK18-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
12241 // CHECK18-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
12242 // CHECK18-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
12243 // CHECK18-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
12244 // CHECK18-NEXT:    store i8* null, i8** [[TMP79]], align 8
12245 // CHECK18-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
12246 // CHECK18-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
12247 // CHECK18-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
12248 // CHECK18-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
12249 // CHECK18-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
12250 // CHECK18-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
12251 // CHECK18-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
12252 // CHECK18-NEXT:    store i8* null, i8** [[TMP84]], align 8
12253 // CHECK18-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
12254 // CHECK18-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
12255 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
12256 // CHECK18-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP85]], i8** [[TMP86]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12257 // CHECK18-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
12258 // CHECK18-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
12259 // CHECK18:       omp_offload.failed20:
12260 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
12261 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT21]]
12262 // CHECK18:       omp_offload.cont21:
12263 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
12264 // CHECK18:       omp_if.else:
12265 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
12266 // CHECK18-NEXT:    br label [[OMP_IF_END]]
12267 // CHECK18:       omp_if.end:
12268 // CHECK18-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
12269 // CHECK18-NEXT:    store i32 [[TMP89]], i32* [[DOTCAPTURE_EXPR_22]], align 4
12270 // CHECK18-NEXT:    [[TMP90:%.*]] = load i32, i32* [[A]], align 4
12271 // CHECK18-NEXT:    [[CONV24:%.*]] = bitcast i64* [[A_CASTED23]] to i32*
12272 // CHECK18-NEXT:    store i32 [[TMP90]], i32* [[CONV24]], align 4
12273 // CHECK18-NEXT:    [[TMP91:%.*]] = load i64, i64* [[A_CASTED23]], align 8
12274 // CHECK18-NEXT:    [[TMP92:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_22]], align 4
12275 // CHECK18-NEXT:    [[CONV26:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED25]] to i32*
12276 // CHECK18-NEXT:    store i32 [[TMP92]], i32* [[CONV26]], align 4
12277 // CHECK18-NEXT:    [[TMP93:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED25]], align 8
12278 // CHECK18-NEXT:    [[TMP94:%.*]] = load i32, i32* [[N_ADDR]], align 4
12279 // CHECK18-NEXT:    [[CMP27:%.*]] = icmp sgt i32 [[TMP94]], 20
12280 // CHECK18-NEXT:    br i1 [[CMP27]], label [[OMP_IF_THEN28:%.*]], label [[OMP_IF_ELSE35:%.*]]
12281 // CHECK18:       omp_if.then28:
12282 // CHECK18-NEXT:    [[TMP95:%.*]] = mul nuw i64 [[TMP2]], 4
12283 // CHECK18-NEXT:    [[TMP96:%.*]] = mul nuw i64 5, [[TMP5]]
12284 // CHECK18-NEXT:    [[TMP97:%.*]] = mul nuw i64 [[TMP96]], 8
12285 // CHECK18-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
12286 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP98]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i64 80, i1 false)
12287 // CHECK18-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
12288 // CHECK18-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i64*
12289 // CHECK18-NEXT:    store i64 [[TMP91]], i64* [[TMP100]], align 8
12290 // CHECK18-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
12291 // CHECK18-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i64*
12292 // CHECK18-NEXT:    store i64 [[TMP91]], i64* [[TMP102]], align 8
12293 // CHECK18-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 0
12294 // CHECK18-NEXT:    store i8* null, i8** [[TMP103]], align 8
12295 // CHECK18-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 1
12296 // CHECK18-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
12297 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 8
12298 // CHECK18-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 1
12299 // CHECK18-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
12300 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 8
12301 // CHECK18-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 1
12302 // CHECK18-NEXT:    store i8* null, i8** [[TMP108]], align 8
12303 // CHECK18-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 2
12304 // CHECK18-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i64*
12305 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP110]], align 8
12306 // CHECK18-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 2
12307 // CHECK18-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i64*
12308 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP112]], align 8
12309 // CHECK18-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 2
12310 // CHECK18-NEXT:    store i8* null, i8** [[TMP113]], align 8
12311 // CHECK18-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 3
12312 // CHECK18-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
12313 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP115]], align 8
12314 // CHECK18-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 3
12315 // CHECK18-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
12316 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP117]], align 8
12317 // CHECK18-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
12318 // CHECK18-NEXT:    store i64 [[TMP95]], i64* [[TMP118]], align 8
12319 // CHECK18-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 3
12320 // CHECK18-NEXT:    store i8* null, i8** [[TMP119]], align 8
12321 // CHECK18-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 4
12322 // CHECK18-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
12323 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 8
12324 // CHECK18-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 4
12325 // CHECK18-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
12326 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 8
12327 // CHECK18-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 4
12328 // CHECK18-NEXT:    store i8* null, i8** [[TMP124]], align 8
12329 // CHECK18-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 5
12330 // CHECK18-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i64*
12331 // CHECK18-NEXT:    store i64 5, i64* [[TMP126]], align 8
12332 // CHECK18-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 5
12333 // CHECK18-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
12334 // CHECK18-NEXT:    store i64 5, i64* [[TMP128]], align 8
12335 // CHECK18-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 5
12336 // CHECK18-NEXT:    store i8* null, i8** [[TMP129]], align 8
12337 // CHECK18-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 6
12338 // CHECK18-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i64*
12339 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP131]], align 8
12340 // CHECK18-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 6
12341 // CHECK18-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i64*
12342 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP133]], align 8
12343 // CHECK18-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 6
12344 // CHECK18-NEXT:    store i8* null, i8** [[TMP134]], align 8
12345 // CHECK18-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 7
12346 // CHECK18-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
12347 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 8
12348 // CHECK18-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 7
12349 // CHECK18-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
12350 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 8
12351 // CHECK18-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
12352 // CHECK18-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 8
12353 // CHECK18-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 7
12354 // CHECK18-NEXT:    store i8* null, i8** [[TMP140]], align 8
12355 // CHECK18-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 8
12356 // CHECK18-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
12357 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 8
12358 // CHECK18-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 8
12359 // CHECK18-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
12360 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 8
12361 // CHECK18-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 8
12362 // CHECK18-NEXT:    store i8* null, i8** [[TMP145]], align 8
12363 // CHECK18-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 9
12364 // CHECK18-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i64*
12365 // CHECK18-NEXT:    store i64 [[TMP93]], i64* [[TMP147]], align 8
12366 // CHECK18-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 9
12367 // CHECK18-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
12368 // CHECK18-NEXT:    store i64 [[TMP93]], i64* [[TMP149]], align 8
12369 // CHECK18-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS31]], i64 0, i64 9
12370 // CHECK18-NEXT:    store i8* null, i8** [[TMP150]], align 8
12371 // CHECK18-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS29]], i32 0, i32 0
12372 // CHECK18-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS30]], i32 0, i32 0
12373 // CHECK18-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
12374 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
12375 // CHECK18-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12376 // CHECK18-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
12377 // CHECK18-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED33:%.*]], label [[OMP_OFFLOAD_CONT34:%.*]]
12378 // CHECK18:       omp_offload.failed33:
12379 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
12380 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT34]]
12381 // CHECK18:       omp_offload.cont34:
12382 // CHECK18-NEXT:    br label [[OMP_IF_END36:%.*]]
12383 // CHECK18:       omp_if.else35:
12384 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i64 [[TMP91]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP93]]) #[[ATTR3]]
12385 // CHECK18-NEXT:    br label [[OMP_IF_END36]]
12386 // CHECK18:       omp_if.end36:
12387 // CHECK18-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
12388 // CHECK18-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
12389 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
12390 // CHECK18-NEXT:    ret i32 [[TMP156]]
12391 //
12392 //
12393 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
12394 // CHECK18-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
12395 // CHECK18-NEXT:  entry:
12396 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12397 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12398 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
12399 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
12400 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
12401 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12402 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12403 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
12404 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12405 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
12406 // CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
12407 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
12408 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
12409 // CHECK18-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
12410 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
12411 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
12412 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
12413 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
12414 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
12415 // CHECK18-NEXT:    ret void
12416 //
12417 //
12418 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
12419 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
12420 // CHECK18-NEXT:  entry:
12421 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12422 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12423 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12424 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12425 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12426 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12427 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12428 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12429 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12430 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
12431 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12432 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12433 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12434 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12435 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12436 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12437 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12438 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12439 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12440 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12441 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12442 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12443 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
12444 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12445 // CHECK18:       cond.true:
12446 // CHECK18-NEXT:    br label [[COND_END:%.*]]
12447 // CHECK18:       cond.false:
12448 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12449 // CHECK18-NEXT:    br label [[COND_END]]
12450 // CHECK18:       cond.end:
12451 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12452 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12453 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12454 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12455 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12456 // CHECK18:       omp.inner.for.cond:
12457 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12458 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12459 // CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12460 // CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12461 // CHECK18:       omp.inner.for.body:
12462 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12463 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
12464 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12465 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12466 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12467 // CHECK18:       omp.body.continue:
12468 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12469 // CHECK18:       omp.inner.for.inc:
12470 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12471 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
12472 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
12473 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
12474 // CHECK18:       omp.inner.for.end:
12475 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12476 // CHECK18:       omp.loop.exit:
12477 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12478 // CHECK18-NEXT:    ret void
12479 //
12480 //
12481 // CHECK18-LABEL: define {{[^@]+}}@.omp_task_privates_map.
12482 // CHECK18-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
12483 // CHECK18-NEXT:  entry:
12484 // CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
12485 // CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
12486 // CHECK18-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
12487 // CHECK18-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
12488 // CHECK18-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
12489 // CHECK18-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
12490 // CHECK18-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
12491 // CHECK18-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
12492 // CHECK18-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
12493 // CHECK18-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
12494 // CHECK18-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
12495 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
12496 // CHECK18-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
12497 // CHECK18-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
12498 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
12499 // CHECK18-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
12500 // CHECK18-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
12501 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
12502 // CHECK18-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
12503 // CHECK18-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
12504 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
12505 // CHECK18-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
12506 // CHECK18-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
12507 // CHECK18-NEXT:    ret void
12508 //
12509 //
12510 // CHECK18-LABEL: define {{[^@]+}}@.omp_task_entry.
12511 // CHECK18-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
12512 // CHECK18-NEXT:  entry:
12513 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
12514 // CHECK18-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
12515 // CHECK18-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
12516 // CHECK18-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
12517 // CHECK18-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
12518 // CHECK18-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
12519 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
12520 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
12521 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
12522 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
12523 // CHECK18-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
12524 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
12525 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
12526 // CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
12527 // CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
12528 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
12529 // CHECK18-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
12530 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
12531 // CHECK18-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
12532 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
12533 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
12534 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
12535 // CHECK18-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
12536 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
12537 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
12538 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
12539 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
12540 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META12:![0-9]+]])
12541 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
12542 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
12543 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
12544 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !21
12545 // CHECK18-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !21
12546 // CHECK18-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
12547 // CHECK18-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
12548 // CHECK18-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !21
12549 // CHECK18-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
12550 // CHECK18-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !21
12551 // CHECK18-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !21
12552 // CHECK18-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !21
12553 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
12554 // CHECK18-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
12555 // CHECK18-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !21
12556 // CHECK18-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !21
12557 // CHECK18-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !21
12558 // CHECK18-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !21
12559 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
12560 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
12561 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
12562 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
12563 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
12564 // CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
12565 // CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
12566 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
12567 // CHECK18-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
12568 // CHECK18-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
12569 // CHECK18-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
12570 // CHECK18:       omp_offload.failed.i:
12571 // CHECK18-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
12572 // CHECK18-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
12573 // CHECK18-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !21
12574 // CHECK18-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !21
12575 // CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
12576 // CHECK18-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
12577 // CHECK18-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !21
12578 // CHECK18-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !21
12579 // CHECK18-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
12580 // CHECK18-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
12581 // CHECK18-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !21
12582 // CHECK18-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !21
12583 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
12584 // CHECK18-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
12585 // CHECK18:       .omp_outlined..1.exit:
12586 // CHECK18-NEXT:    ret i32 0
12587 //
12588 //
12589 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
12590 // CHECK18-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
12591 // CHECK18-NEXT:  entry:
12592 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12593 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
12594 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12595 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12596 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
12597 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
12598 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
12599 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
12600 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
12601 // CHECK18-NEXT:    ret void
12602 //
12603 //
12604 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
12605 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
12606 // CHECK18-NEXT:  entry:
12607 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12608 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12609 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12610 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12611 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12612 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12613 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12614 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12615 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12616 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
12617 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12618 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12619 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12620 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12621 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12622 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12623 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12624 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12625 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12626 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12627 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12628 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12629 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
12630 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12631 // CHECK18:       cond.true:
12632 // CHECK18-NEXT:    br label [[COND_END:%.*]]
12633 // CHECK18:       cond.false:
12634 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12635 // CHECK18-NEXT:    br label [[COND_END]]
12636 // CHECK18:       cond.end:
12637 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12638 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12639 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12640 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12641 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12642 // CHECK18:       omp.inner.for.cond:
12643 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12644 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12645 // CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12646 // CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12647 // CHECK18:       omp.inner.for.body:
12648 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12649 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
12650 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12651 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12652 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
12653 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
12654 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[CONV]], align 4
12655 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12656 // CHECK18:       omp.body.continue:
12657 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12658 // CHECK18:       omp.inner.for.inc:
12659 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12660 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
12661 // CHECK18-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
12662 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
12663 // CHECK18:       omp.inner.for.end:
12664 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12665 // CHECK18:       omp.loop.exit:
12666 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12667 // CHECK18-NEXT:    ret void
12668 //
12669 //
12670 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
12671 // CHECK18-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
12672 // CHECK18-NEXT:  entry:
12673 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12674 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
12675 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12676 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12677 // CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
12678 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
12679 // CHECK18-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
12680 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
12681 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
12682 // CHECK18-NEXT:    ret void
12683 //
12684 //
12685 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
12686 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
12687 // CHECK18-NEXT:  entry:
12688 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12689 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12690 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12691 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12692 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12693 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12694 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12695 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12696 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12697 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
12698 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12699 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12700 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12701 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12702 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12703 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12704 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12705 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12706 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12707 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12708 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12709 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12710 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
12711 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12712 // CHECK18:       cond.true:
12713 // CHECK18-NEXT:    br label [[COND_END:%.*]]
12714 // CHECK18:       cond.false:
12715 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12716 // CHECK18-NEXT:    br label [[COND_END]]
12717 // CHECK18:       cond.end:
12718 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12719 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12720 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12721 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12722 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12723 // CHECK18:       omp.inner.for.cond:
12724 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12725 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12726 // CHECK18-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12727 // CHECK18-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12728 // CHECK18:       omp.inner.for.body:
12729 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12730 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
12731 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12732 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12733 // CHECK18-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
12734 // CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
12735 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
12736 // CHECK18-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
12737 // CHECK18-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
12738 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12739 // CHECK18:       omp.body.continue:
12740 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12741 // CHECK18:       omp.inner.for.inc:
12742 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12743 // CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
12744 // CHECK18-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
12745 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
12746 // CHECK18:       omp.inner.for.end:
12747 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12748 // CHECK18:       omp.loop.exit:
12749 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12750 // CHECK18-NEXT:    ret void
12751 //
12752 //
12753 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
12754 // CHECK18-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
12755 // CHECK18-NEXT:  entry:
12756 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12757 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12758 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
12759 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
12760 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12761 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12762 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12763 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12764 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
12765 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
12766 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
12767 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
12768 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
12769 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
12770 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
12771 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
12772 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
12773 // CHECK18-NEXT:    ret void
12774 //
12775 //
12776 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..6
12777 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
12778 // CHECK18-NEXT:  entry:
12779 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12780 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12781 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12782 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
12783 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12784 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12785 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12786 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12787 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12788 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12789 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
12790 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12791 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12792 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12793 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
12794 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12795 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
12796 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12797 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12798 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12799 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12800 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12801 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12802 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12803 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12804 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
12805 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12806 // CHECK18:       cond.true:
12807 // CHECK18-NEXT:    br label [[COND_END:%.*]]
12808 // CHECK18:       cond.false:
12809 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12810 // CHECK18-NEXT:    br label [[COND_END]]
12811 // CHECK18:       cond.end:
12812 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12813 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12814 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12815 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12816 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12817 // CHECK18:       omp.inner.for.cond:
12818 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12819 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12820 // CHECK18-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12821 // CHECK18-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12822 // CHECK18:       omp.inner.for.body:
12823 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12824 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
12825 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12826 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12827 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
12828 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
12829 // CHECK18-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
12830 // CHECK18-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
12831 // CHECK18-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
12832 // CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
12833 // CHECK18-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
12834 // CHECK18-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
12835 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12836 // CHECK18:       omp.body.continue:
12837 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12838 // CHECK18:       omp.inner.for.inc:
12839 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12840 // CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
12841 // CHECK18-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
12842 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
12843 // CHECK18:       omp.inner.for.end:
12844 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12845 // CHECK18:       omp.loop.exit:
12846 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12847 // CHECK18-NEXT:    ret void
12848 //
12849 //
12850 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
12851 // CHECK18-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
12852 // CHECK18-NEXT:  entry:
12853 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12854 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
12855 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
12856 // CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
12857 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
12858 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
12859 // CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
12860 // CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
12861 // CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
12862 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12863 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
12864 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
12865 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12866 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
12867 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
12868 // CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
12869 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
12870 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
12871 // CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
12872 // CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
12873 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
12874 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12875 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12876 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
12877 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
12878 // CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
12879 // CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
12880 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
12881 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
12882 // CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
12883 // CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
12884 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
12885 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
12886 // CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
12887 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
12888 // CHECK18-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
12889 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
12890 // CHECK18-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
12891 // CHECK18-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
12892 // CHECK18-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
12893 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
12894 // CHECK18-NEXT:    ret void
12895 //
12896 //
12897 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
12898 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
12899 // CHECK18-NEXT:  entry:
12900 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12901 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12902 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
12903 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
12904 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
12905 // CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
12906 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
12907 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
12908 // CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
12909 // CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
12910 // CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
12911 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12912 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12913 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12914 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12915 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12916 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12917 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12918 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
12919 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12920 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12921 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
12922 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
12923 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
12924 // CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
12925 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
12926 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
12927 // CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
12928 // CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
12929 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
12930 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12931 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
12932 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
12933 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
12934 // CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
12935 // CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
12936 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
12937 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
12938 // CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
12939 // CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
12940 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
12941 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12942 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
12943 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12944 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12945 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
12946 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12947 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
12948 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
12949 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
12950 // CHECK18:       omp.dispatch.cond:
12951 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12952 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
12953 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12954 // CHECK18:       cond.true:
12955 // CHECK18-NEXT:    br label [[COND_END:%.*]]
12956 // CHECK18:       cond.false:
12957 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12958 // CHECK18-NEXT:    br label [[COND_END]]
12959 // CHECK18:       cond.end:
12960 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
12961 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12962 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12963 // CHECK18-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
12964 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12965 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12966 // CHECK18-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
12967 // CHECK18-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
12968 // CHECK18:       omp.dispatch.body:
12969 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12970 // CHECK18:       omp.inner.for.cond:
12971 // CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
12972 // CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
12973 // CHECK18-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
12974 // CHECK18-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12975 // CHECK18:       omp.inner.for.body:
12976 // CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
12977 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
12978 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12979 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
12980 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !22
12981 // CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
12982 // CHECK18-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !22
12983 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
12984 // CHECK18-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !22
12985 // CHECK18-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
12986 // CHECK18-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
12987 // CHECK18-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
12988 // CHECK18-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !22
12989 // CHECK18-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
12990 // CHECK18-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
12991 // CHECK18-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
12992 // CHECK18-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
12993 // CHECK18-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
12994 // CHECK18-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !22
12995 // CHECK18-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
12996 // CHECK18-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
12997 // CHECK18-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
12998 // CHECK18-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
12999 // CHECK18-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !22
13000 // CHECK18-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
13001 // CHECK18-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
13002 // CHECK18-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
13003 // CHECK18-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
13004 // CHECK18-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
13005 // CHECK18-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !22
13006 // CHECK18-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
13007 // CHECK18-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !22
13008 // CHECK18-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
13009 // CHECK18-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !22
13010 // CHECK18-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
13011 // CHECK18-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !22
13012 // CHECK18-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
13013 // CHECK18-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
13014 // CHECK18-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
13015 // CHECK18-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !22
13016 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13017 // CHECK18:       omp.body.continue:
13018 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13019 // CHECK18:       omp.inner.for.inc:
13020 // CHECK18-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13021 // CHECK18-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
13022 // CHECK18-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13023 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
13024 // CHECK18:       omp.inner.for.end:
13025 // CHECK18-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
13026 // CHECK18:       omp.dispatch.inc:
13027 // CHECK18-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13028 // CHECK18-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13029 // CHECK18-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
13030 // CHECK18-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
13031 // CHECK18-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13032 // CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13033 // CHECK18-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
13034 // CHECK18-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
13035 // CHECK18-NEXT:    br label [[OMP_DISPATCH_COND]]
13036 // CHECK18:       omp.dispatch.end:
13037 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
13038 // CHECK18-NEXT:    ret void
13039 //
13040 //
13041 // CHECK18-LABEL: define {{[^@]+}}@_Z3bari
13042 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
13043 // CHECK18-NEXT:  entry:
13044 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
13045 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
13046 // CHECK18-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
13047 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
13048 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
13049 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
13050 // CHECK18-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
13051 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
13052 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
13053 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
13054 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
13055 // CHECK18-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
13056 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
13057 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
13058 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
13059 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
13060 // CHECK18-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
13061 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
13062 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
13063 // CHECK18-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
13064 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
13065 // CHECK18-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
13066 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
13067 // CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
13068 // CHECK18-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
13069 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
13070 // CHECK18-NEXT:    ret i32 [[TMP8]]
13071 //
13072 //
13073 // CHECK18-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
13074 // CHECK18-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
13075 // CHECK18-NEXT:  entry:
13076 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
13077 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
13078 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
13079 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
13080 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
13081 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
13082 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
13083 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
13084 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
13085 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
13086 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13087 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
13088 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
13089 // CHECK18-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
13090 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
13091 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
13092 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
13093 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
13094 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
13095 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
13096 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
13097 // CHECK18-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
13098 // CHECK18-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
13099 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
13100 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
13101 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
13102 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
13103 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
13104 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
13105 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
13106 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
13107 // CHECK18:       omp_if.then:
13108 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
13109 // CHECK18-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
13110 // CHECK18-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
13111 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
13112 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i64 40, i1 false)
13113 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13114 // CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
13115 // CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
13116 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13117 // CHECK18-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
13118 // CHECK18-NEXT:    store double* [[A]], double** [[TMP14]], align 8
13119 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
13120 // CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
13121 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
13122 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
13123 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
13124 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
13125 // CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
13126 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
13127 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
13128 // CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
13129 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
13130 // CHECK18-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
13131 // CHECK18-NEXT:    store i64 2, i64* [[TMP22]], align 8
13132 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
13133 // CHECK18-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
13134 // CHECK18-NEXT:    store i64 2, i64* [[TMP24]], align 8
13135 // CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
13136 // CHECK18-NEXT:    store i8* null, i8** [[TMP25]], align 8
13137 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
13138 // CHECK18-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
13139 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
13140 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
13141 // CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
13142 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
13143 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
13144 // CHECK18-NEXT:    store i8* null, i8** [[TMP30]], align 8
13145 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
13146 // CHECK18-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
13147 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
13148 // CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
13149 // CHECK18-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
13150 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
13151 // CHECK18-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
13152 // CHECK18-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
13153 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
13154 // CHECK18-NEXT:    store i8* null, i8** [[TMP36]], align 8
13155 // CHECK18-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13156 // CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13157 // CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
13158 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
13159 // CHECK18-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13160 // CHECK18-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
13161 // CHECK18-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
13162 // CHECK18:       omp_offload.failed:
13163 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
13164 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
13165 // CHECK18:       omp_offload.cont:
13166 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
13167 // CHECK18:       omp_if.else:
13168 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
13169 // CHECK18-NEXT:    br label [[OMP_IF_END]]
13170 // CHECK18:       omp_if.end:
13171 // CHECK18-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
13172 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
13173 // CHECK18-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
13174 // CHECK18-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
13175 // CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
13176 // CHECK18-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
13177 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
13178 // CHECK18-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
13179 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
13180 // CHECK18-NEXT:    ret i32 [[ADD4]]
13181 //
13182 //
13183 // CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
13184 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
13185 // CHECK18-NEXT:  entry:
13186 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
13187 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
13188 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
13189 // CHECK18-NEXT:    [[AAA:%.*]] = alloca i8, align 1
13190 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
13191 // CHECK18-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
13192 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13193 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13194 // CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
13195 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
13196 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
13197 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
13198 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13199 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13200 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
13201 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
13202 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
13203 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
13204 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
13205 // CHECK18-NEXT:    store i8 0, i8* [[AAA]], align 1
13206 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
13207 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
13208 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
13209 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[N_CASTED]], align 8
13210 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
13211 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13212 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[CONV1]], align 4
13213 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[A_CASTED]], align 8
13214 // CHECK18-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
13215 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13216 // CHECK18-NEXT:    store i16 [[TMP4]], i16* [[CONV2]], align 2
13217 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13218 // CHECK18-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
13219 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
13220 // CHECK18-NEXT:    store i8 [[TMP6]], i8* [[CONV3]], align 1
13221 // CHECK18-NEXT:    [[TMP7:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
13222 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
13223 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
13224 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
13225 // CHECK18:       omp_if.then:
13226 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13227 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
13228 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
13229 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13230 // CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
13231 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP12]], align 8
13232 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
13233 // CHECK18-NEXT:    store i8* null, i8** [[TMP13]], align 8
13234 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
13235 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
13236 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
13237 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
13238 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
13239 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP17]], align 8
13240 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
13241 // CHECK18-NEXT:    store i8* null, i8** [[TMP18]], align 8
13242 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
13243 // CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
13244 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
13245 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
13246 // CHECK18-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
13247 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP22]], align 8
13248 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
13249 // CHECK18-NEXT:    store i8* null, i8** [[TMP23]], align 8
13250 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
13251 // CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
13252 // CHECK18-NEXT:    store i64 [[TMP7]], i64* [[TMP25]], align 8
13253 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
13254 // CHECK18-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
13255 // CHECK18-NEXT:    store i64 [[TMP7]], i64* [[TMP27]], align 8
13256 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
13257 // CHECK18-NEXT:    store i8* null, i8** [[TMP28]], align 8
13258 // CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
13259 // CHECK18-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
13260 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 8
13261 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
13262 // CHECK18-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
13263 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 8
13264 // CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
13265 // CHECK18-NEXT:    store i8* null, i8** [[TMP33]], align 8
13266 // CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13267 // CHECK18-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13268 // CHECK18-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
13269 // CHECK18-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
13270 // CHECK18-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
13271 // CHECK18-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_4]], align 4
13272 // CHECK18-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
13273 // CHECK18-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13274 // CHECK18-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
13275 // CHECK18-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
13276 // CHECK18-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
13277 // CHECK18-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
13278 // CHECK18-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
13279 // CHECK18-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
13280 // CHECK18-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
13281 // CHECK18-NEXT:    [[ADD8:%.*]] = add i32 [[TMP40]], 1
13282 // CHECK18-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD8]] to i64
13283 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
13284 // CHECK18-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13285 // CHECK18-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
13286 // CHECK18-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
13287 // CHECK18:       omp_offload.failed:
13288 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
13289 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
13290 // CHECK18:       omp_offload.cont:
13291 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
13292 // CHECK18:       omp_if.else:
13293 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], i64 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
13294 // CHECK18-NEXT:    br label [[OMP_IF_END]]
13295 // CHECK18:       omp_if.end:
13296 // CHECK18-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
13297 // CHECK18-NEXT:    ret i32 [[TMP44]]
13298 //
13299 //
13300 // CHECK18-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
13301 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
13302 // CHECK18-NEXT:  entry:
13303 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
13304 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
13305 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
13306 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
13307 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13308 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13309 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
13310 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
13311 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
13312 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13313 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
13314 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
13315 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
13316 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
13317 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13318 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
13319 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
13320 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
13321 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13322 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
13323 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13324 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
13325 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
13326 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
13327 // CHECK18:       omp_if.then:
13328 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13329 // CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
13330 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
13331 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13332 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
13333 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
13334 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
13335 // CHECK18-NEXT:    store i8* null, i8** [[TMP9]], align 8
13336 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
13337 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
13338 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
13339 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
13340 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
13341 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
13342 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
13343 // CHECK18-NEXT:    store i8* null, i8** [[TMP14]], align 8
13344 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
13345 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
13346 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
13347 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
13348 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
13349 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
13350 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
13351 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
13352 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13353 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13354 // CHECK18-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
13355 // CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13356 // CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
13357 // CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
13358 // CHECK18:       omp_offload.failed:
13359 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
13360 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
13361 // CHECK18:       omp_offload.cont:
13362 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
13363 // CHECK18:       omp_if.else:
13364 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
13365 // CHECK18-NEXT:    br label [[OMP_IF_END]]
13366 // CHECK18:       omp_if.end:
13367 // CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
13368 // CHECK18-NEXT:    ret i32 [[TMP24]]
13369 //
13370 //
13371 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
13372 // CHECK18-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
13373 // CHECK18-NEXT:  entry:
13374 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
13375 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
13376 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13377 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13378 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
13379 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
13380 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
13381 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
13382 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13383 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13384 // CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
13385 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
13386 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
13387 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13388 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13389 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
13390 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
13391 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
13392 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
13393 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
13394 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
13395 // CHECK18-NEXT:    ret void
13396 //
13397 //
13398 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..12
13399 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
13400 // CHECK18-NEXT:  entry:
13401 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13402 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13403 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
13404 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
13405 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13406 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13407 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
13408 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13409 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13410 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13411 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13412 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13413 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13414 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
13415 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13416 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13417 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
13418 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
13419 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13420 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13421 // CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
13422 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
13423 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
13424 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13425 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13426 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
13427 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13428 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
13429 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13430 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13431 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13432 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
13433 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13434 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13435 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
13436 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13437 // CHECK18:       cond.true:
13438 // CHECK18-NEXT:    br label [[COND_END:%.*]]
13439 // CHECK18:       cond.false:
13440 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13441 // CHECK18-NEXT:    br label [[COND_END]]
13442 // CHECK18:       cond.end:
13443 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
13444 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13445 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13446 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
13447 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13448 // CHECK18:       omp.inner.for.cond:
13449 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13450 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13451 // CHECK18-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
13452 // CHECK18-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13453 // CHECK18:       omp.inner.for.body:
13454 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13455 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
13456 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13457 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
13458 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
13459 // CHECK18-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
13460 // CHECK18-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
13461 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
13462 // CHECK18-NEXT:    store double [[ADD5]], double* [[A]], align 8
13463 // CHECK18-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
13464 // CHECK18-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
13465 // CHECK18-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
13466 // CHECK18-NEXT:    store double [[INC]], double* [[A6]], align 8
13467 // CHECK18-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
13468 // CHECK18-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
13469 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
13470 // CHECK18-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
13471 // CHECK18-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
13472 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13473 // CHECK18:       omp.body.continue:
13474 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13475 // CHECK18:       omp.inner.for.inc:
13476 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13477 // CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
13478 // CHECK18-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
13479 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
13480 // CHECK18:       omp.inner.for.end:
13481 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13482 // CHECK18:       omp.loop.exit:
13483 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
13484 // CHECK18-NEXT:    ret void
13485 //
13486 //
13487 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
13488 // CHECK18-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13489 // CHECK18-NEXT:  entry:
13490 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
13491 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13492 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13493 // CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
13494 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13495 // CHECK18-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
13496 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13497 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13498 // CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
13499 // CHECK18-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
13500 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13501 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13502 // CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
13503 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13504 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
13505 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13506 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13507 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
13508 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13509 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13510 // CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
13511 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
13512 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
13513 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
13514 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13515 // CHECK18-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
13516 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
13517 // CHECK18-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
13518 // CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13519 // CHECK18-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
13520 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13521 // CHECK18-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
13522 // CHECK18-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
13523 // CHECK18-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
13524 // CHECK18-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
13525 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
13526 // CHECK18-NEXT:    ret void
13527 //
13528 //
13529 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..15
13530 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13531 // CHECK18-NEXT:  entry:
13532 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13533 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13534 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
13535 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13536 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13537 // CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
13538 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13539 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13540 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13541 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13542 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
13543 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
13544 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
13545 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13546 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13547 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13548 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13549 // CHECK18-NEXT:    [[I8:%.*]] = alloca i32, align 4
13550 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13551 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13552 // CHECK18-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
13553 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13554 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13555 // CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
13556 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13557 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
13558 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13559 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13560 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
13561 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13562 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
13563 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
13564 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
13565 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
13566 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
13567 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13568 // CHECK18-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
13569 // CHECK18-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
13570 // CHECK18-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
13571 // CHECK18-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
13572 // CHECK18-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
13573 // CHECK18-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
13574 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13575 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
13576 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13577 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
13578 // CHECK18-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
13579 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13580 // CHECK18:       omp.precond.then:
13581 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13582 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
13583 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
13584 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13585 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13586 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13587 // CHECK18-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
13588 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13589 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13590 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
13591 // CHECK18-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
13592 // CHECK18-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13593 // CHECK18:       cond.true:
13594 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
13595 // CHECK18-NEXT:    br label [[COND_END:%.*]]
13596 // CHECK18:       cond.false:
13597 // CHECK18-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13598 // CHECK18-NEXT:    br label [[COND_END]]
13599 // CHECK18:       cond.end:
13600 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
13601 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13602 // CHECK18-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13603 // CHECK18-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
13604 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13605 // CHECK18:       omp.inner.for.cond:
13606 // CHECK18-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13607 // CHECK18-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13608 // CHECK18-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
13609 // CHECK18-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
13610 // CHECK18-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13611 // CHECK18:       omp.inner.for.body:
13612 // CHECK18-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13613 // CHECK18-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13614 // CHECK18-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
13615 // CHECK18-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
13616 // CHECK18-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
13617 // CHECK18-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
13618 // CHECK18-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
13619 // CHECK18-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
13620 // CHECK18-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
13621 // CHECK18-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
13622 // CHECK18-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
13623 // CHECK18-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
13624 // CHECK18-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
13625 // CHECK18-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
13626 // CHECK18-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
13627 // CHECK18-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
13628 // CHECK18-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
13629 // CHECK18-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
13630 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
13631 // CHECK18-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13632 // CHECK18-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
13633 // CHECK18-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
13634 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13635 // CHECK18:       omp.body.continue:
13636 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13637 // CHECK18:       omp.inner.for.inc:
13638 // CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13639 // CHECK18-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
13640 // CHECK18-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
13641 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
13642 // CHECK18:       omp.inner.for.end:
13643 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13644 // CHECK18:       omp.loop.exit:
13645 // CHECK18-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13646 // CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
13647 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
13648 // CHECK18-NEXT:    br label [[OMP_PRECOND_END]]
13649 // CHECK18:       omp.precond.end:
13650 // CHECK18-NEXT:    ret void
13651 //
13652 //
13653 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
13654 // CHECK18-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13655 // CHECK18-NEXT:  entry:
13656 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13657 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13658 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13659 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13660 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13661 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13662 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13663 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13664 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13665 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13666 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13667 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13668 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13669 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
13670 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
13671 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
13672 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13673 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
13674 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13675 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
13676 // CHECK18-NEXT:    ret void
13677 //
13678 //
13679 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..18
13680 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13681 // CHECK18-NEXT:  entry:
13682 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13683 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13684 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13685 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13686 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13687 // CHECK18-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13688 // CHECK18-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13689 // CHECK18-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13690 // CHECK18-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13691 // CHECK18-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13692 // CHECK18-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13693 // CHECK18-NEXT:    [[I:%.*]] = alloca i32, align 4
13694 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13695 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13696 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13697 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13698 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13699 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13700 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13701 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13702 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13703 // CHECK18-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
13704 // CHECK18-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13705 // CHECK18-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13706 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13707 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
13708 // CHECK18-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13709 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13710 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
13711 // CHECK18-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13712 // CHECK18:       cond.true:
13713 // CHECK18-NEXT:    br label [[COND_END:%.*]]
13714 // CHECK18:       cond.false:
13715 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13716 // CHECK18-NEXT:    br label [[COND_END]]
13717 // CHECK18:       cond.end:
13718 // CHECK18-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
13719 // CHECK18-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13720 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13721 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
13722 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13723 // CHECK18:       omp.inner.for.cond:
13724 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13725 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13726 // CHECK18-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13727 // CHECK18-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13728 // CHECK18:       omp.inner.for.body:
13729 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13730 // CHECK18-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
13731 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13732 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
13733 // CHECK18-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
13734 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
13735 // CHECK18-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
13736 // CHECK18-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
13737 // CHECK18-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
13738 // CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
13739 // CHECK18-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
13740 // CHECK18-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
13741 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
13742 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13743 // CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
13744 // CHECK18-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
13745 // CHECK18-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13746 // CHECK18:       omp.body.continue:
13747 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13748 // CHECK18:       omp.inner.for.inc:
13749 // CHECK18-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13750 // CHECK18-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
13751 // CHECK18-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
13752 // CHECK18-NEXT:    br label [[OMP_INNER_FOR_COND]]
13753 // CHECK18:       omp.inner.for.end:
13754 // CHECK18-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13755 // CHECK18:       omp.loop.exit:
13756 // CHECK18-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
13757 // CHECK18-NEXT:    ret void
13758 //
13759 //
13760 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
13761 // CHECK18-SAME: () #[[ATTR4]] {
13762 // CHECK18-NEXT:  entry:
13763 // CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
13764 // CHECK18-NEXT:    ret void
13765 //
13766 //
13767 // CHECK19-LABEL: define {{[^@]+}}@_Z3fooi
13768 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
13769 // CHECK19-NEXT:  entry:
13770 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
13771 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
13772 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
13773 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
13774 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
13775 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
13776 // CHECK19-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
13777 // CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
13778 // CHECK19-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
13779 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13780 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
13781 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
13782 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
13783 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
13784 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
13785 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
13786 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
13787 // CHECK19-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
13788 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
13789 // CHECK19-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
13790 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
13791 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
13792 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
13793 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13794 // CHECK19-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
13795 // CHECK19-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
13796 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
13797 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
13798 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
13799 // CHECK19-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
13800 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
13801 // CHECK19-NEXT:    [[A_CASTED19:%.*]] = alloca i32, align 4
13802 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED20:%.*]] = alloca i32, align 4
13803 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [10 x i8*], align 4
13804 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [10 x i8*], align 4
13805 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [10 x i8*], align 4
13806 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
13807 // CHECK19-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
13808 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
13809 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
13810 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
13811 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
13812 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
13813 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
13814 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
13815 // CHECK19-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
13816 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
13817 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
13818 // CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
13819 // CHECK19-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
13820 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
13821 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
13822 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
13823 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
13824 // CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
13825 // CHECK19-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
13826 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
13827 // CHECK19-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
13828 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
13829 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13830 // CHECK19-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13831 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13832 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13833 // CHECK19-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
13834 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
13835 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13836 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
13837 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
13838 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13839 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
13840 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
13841 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
13842 // CHECK19-NEXT:    store i8* null, i8** [[TMP17]], align 4
13843 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
13844 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
13845 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
13846 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
13847 // CHECK19-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
13848 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
13849 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
13850 // CHECK19-NEXT:    store i8* null, i8** [[TMP22]], align 4
13851 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
13852 // CHECK19-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
13853 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
13854 // CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
13855 // CHECK19-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
13856 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
13857 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
13858 // CHECK19-NEXT:    store i8* null, i8** [[TMP27]], align 4
13859 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13860 // CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13861 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
13862 // CHECK19-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
13863 // CHECK19-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
13864 // CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
13865 // CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13866 // CHECK19-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
13867 // CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
13868 // CHECK19-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13869 // CHECK19-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
13870 // CHECK19-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 72, i32 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
13871 // CHECK19-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
13872 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
13873 // CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
13874 // CHECK19-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
13875 // CHECK19-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
13876 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
13877 // CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
13878 // CHECK19-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
13879 // CHECK19-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
13880 // CHECK19-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
13881 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
13882 // CHECK19-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
13883 // CHECK19-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
13884 // CHECK19-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
13885 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
13886 // CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
13887 // CHECK19-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
13888 // CHECK19-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
13889 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
13890 // CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
13891 // CHECK19-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
13892 // CHECK19-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
13893 // CHECK19-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP36]])
13894 // CHECK19-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
13895 // CHECK19-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
13896 // CHECK19-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
13897 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
13898 // CHECK19-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
13899 // CHECK19-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
13900 // CHECK19-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
13901 // CHECK19-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
13902 // CHECK19-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
13903 // CHECK19-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
13904 // CHECK19-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
13905 // CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
13906 // CHECK19-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
13907 // CHECK19-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
13908 // CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
13909 // CHECK19-NEXT:    store i8* null, i8** [[TMP63]], align 4
13910 // CHECK19-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
13911 // CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
13912 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
13913 // CHECK19-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13914 // CHECK19-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
13915 // CHECK19-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
13916 // CHECK19:       omp_offload.failed:
13917 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i32 [[TMP58]]) #[[ATTR3]]
13918 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
13919 // CHECK19:       omp_offload.cont:
13920 // CHECK19-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
13921 // CHECK19-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
13922 // CHECK19-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
13923 // CHECK19-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
13924 // CHECK19-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
13925 // CHECK19-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
13926 // CHECK19-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
13927 // CHECK19-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
13928 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
13929 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
13930 // CHECK19:       omp_if.then:
13931 // CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
13932 // CHECK19-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
13933 // CHECK19-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
13934 // CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
13935 // CHECK19-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
13936 // CHECK19-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
13937 // CHECK19-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
13938 // CHECK19-NEXT:    store i8* null, i8** [[TMP77]], align 4
13939 // CHECK19-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
13940 // CHECK19-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
13941 // CHECK19-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
13942 // CHECK19-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
13943 // CHECK19-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
13944 // CHECK19-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
13945 // CHECK19-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
13946 // CHECK19-NEXT:    store i8* null, i8** [[TMP82]], align 4
13947 // CHECK19-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
13948 // CHECK19-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
13949 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
13950 // CHECK19-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP83]], i8** [[TMP84]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13951 // CHECK19-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
13952 // CHECK19-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
13953 // CHECK19:       omp_offload.failed16:
13954 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
13955 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
13956 // CHECK19:       omp_offload.cont17:
13957 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
13958 // CHECK19:       omp_if.else:
13959 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
13960 // CHECK19-NEXT:    br label [[OMP_IF_END]]
13961 // CHECK19:       omp_if.end:
13962 // CHECK19-NEXT:    [[TMP87:%.*]] = load i32, i32* [[N_ADDR]], align 4
13963 // CHECK19-NEXT:    store i32 [[TMP87]], i32* [[DOTCAPTURE_EXPR_18]], align 4
13964 // CHECK19-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A]], align 4
13965 // CHECK19-NEXT:    store i32 [[TMP88]], i32* [[A_CASTED19]], align 4
13966 // CHECK19-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A_CASTED19]], align 4
13967 // CHECK19-NEXT:    [[TMP90:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
13968 // CHECK19-NEXT:    store i32 [[TMP90]], i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
13969 // CHECK19-NEXT:    [[TMP91:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
13970 // CHECK19-NEXT:    [[TMP92:%.*]] = load i32, i32* [[N_ADDR]], align 4
13971 // CHECK19-NEXT:    [[CMP21:%.*]] = icmp sgt i32 [[TMP92]], 20
13972 // CHECK19-NEXT:    br i1 [[CMP21]], label [[OMP_IF_THEN22:%.*]], label [[OMP_IF_ELSE29:%.*]]
13973 // CHECK19:       omp_if.then22:
13974 // CHECK19-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP1]], 4
13975 // CHECK19-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
13976 // CHECK19-NEXT:    [[TMP95:%.*]] = mul nuw i32 5, [[TMP3]]
13977 // CHECK19-NEXT:    [[TMP96:%.*]] = mul nuw i32 [[TMP95]], 8
13978 // CHECK19-NEXT:    [[TMP97:%.*]] = sext i32 [[TMP96]] to i64
13979 // CHECK19-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
13980 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP98]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i32 80, i1 false)
13981 // CHECK19-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
13982 // CHECK19-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
13983 // CHECK19-NEXT:    store i32 [[TMP89]], i32* [[TMP100]], align 4
13984 // CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
13985 // CHECK19-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i32*
13986 // CHECK19-NEXT:    store i32 [[TMP89]], i32* [[TMP102]], align 4
13987 // CHECK19-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
13988 // CHECK19-NEXT:    store i8* null, i8** [[TMP103]], align 4
13989 // CHECK19-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
13990 // CHECK19-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
13991 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 4
13992 // CHECK19-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
13993 // CHECK19-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
13994 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 4
13995 // CHECK19-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
13996 // CHECK19-NEXT:    store i8* null, i8** [[TMP108]], align 4
13997 // CHECK19-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 2
13998 // CHECK19-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
13999 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP110]], align 4
14000 // CHECK19-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 2
14001 // CHECK19-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i32*
14002 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP112]], align 4
14003 // CHECK19-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 2
14004 // CHECK19-NEXT:    store i8* null, i8** [[TMP113]], align 4
14005 // CHECK19-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 3
14006 // CHECK19-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
14007 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP115]], align 4
14008 // CHECK19-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 3
14009 // CHECK19-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
14010 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP117]], align 4
14011 // CHECK19-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
14012 // CHECK19-NEXT:    store i64 [[TMP94]], i64* [[TMP118]], align 4
14013 // CHECK19-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 3
14014 // CHECK19-NEXT:    store i8* null, i8** [[TMP119]], align 4
14015 // CHECK19-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 4
14016 // CHECK19-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
14017 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 4
14018 // CHECK19-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 4
14019 // CHECK19-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
14020 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 4
14021 // CHECK19-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 4
14022 // CHECK19-NEXT:    store i8* null, i8** [[TMP124]], align 4
14023 // CHECK19-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 5
14024 // CHECK19-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i32*
14025 // CHECK19-NEXT:    store i32 5, i32* [[TMP126]], align 4
14026 // CHECK19-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 5
14027 // CHECK19-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
14028 // CHECK19-NEXT:    store i32 5, i32* [[TMP128]], align 4
14029 // CHECK19-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 5
14030 // CHECK19-NEXT:    store i8* null, i8** [[TMP129]], align 4
14031 // CHECK19-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 6
14032 // CHECK19-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i32*
14033 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP131]], align 4
14034 // CHECK19-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 6
14035 // CHECK19-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i32*
14036 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP133]], align 4
14037 // CHECK19-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 6
14038 // CHECK19-NEXT:    store i8* null, i8** [[TMP134]], align 4
14039 // CHECK19-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 7
14040 // CHECK19-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
14041 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 4
14042 // CHECK19-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 7
14043 // CHECK19-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
14044 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 4
14045 // CHECK19-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
14046 // CHECK19-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 4
14047 // CHECK19-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 7
14048 // CHECK19-NEXT:    store i8* null, i8** [[TMP140]], align 4
14049 // CHECK19-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 8
14050 // CHECK19-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
14051 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 4
14052 // CHECK19-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 8
14053 // CHECK19-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
14054 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 4
14055 // CHECK19-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 8
14056 // CHECK19-NEXT:    store i8* null, i8** [[TMP145]], align 4
14057 // CHECK19-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 9
14058 // CHECK19-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
14059 // CHECK19-NEXT:    store i32 [[TMP91]], i32* [[TMP147]], align 4
14060 // CHECK19-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 9
14061 // CHECK19-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
14062 // CHECK19-NEXT:    store i32 [[TMP91]], i32* [[TMP149]], align 4
14063 // CHECK19-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 9
14064 // CHECK19-NEXT:    store i8* null, i8** [[TMP150]], align 4
14065 // CHECK19-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
14066 // CHECK19-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
14067 // CHECK19-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
14068 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
14069 // CHECK19-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14070 // CHECK19-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
14071 // CHECK19-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
14072 // CHECK19:       omp_offload.failed27:
14073 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
14074 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
14075 // CHECK19:       omp_offload.cont28:
14076 // CHECK19-NEXT:    br label [[OMP_IF_END30:%.*]]
14077 // CHECK19:       omp_if.else29:
14078 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
14079 // CHECK19-NEXT:    br label [[OMP_IF_END30]]
14080 // CHECK19:       omp_if.end30:
14081 // CHECK19-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
14082 // CHECK19-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
14083 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
14084 // CHECK19-NEXT:    ret i32 [[TMP156]]
14085 //
14086 //
14087 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
14088 // CHECK19-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
14089 // CHECK19-NEXT:  entry:
14090 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14091 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14092 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
14093 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14094 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
14095 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14096 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14097 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14098 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14099 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14100 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14101 // CHECK19-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
14102 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
14103 // CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14104 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
14105 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14106 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
14107 // CHECK19-NEXT:    ret void
14108 //
14109 //
14110 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
14111 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
14112 // CHECK19-NEXT:  entry:
14113 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14114 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14115 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14116 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14117 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14118 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14119 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14120 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14121 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14122 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
14123 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14124 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14125 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14126 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14127 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14128 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
14129 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14130 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14131 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14132 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
14133 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14134 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14135 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
14136 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14137 // CHECK19:       cond.true:
14138 // CHECK19-NEXT:    br label [[COND_END:%.*]]
14139 // CHECK19:       cond.false:
14140 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14141 // CHECK19-NEXT:    br label [[COND_END]]
14142 // CHECK19:       cond.end:
14143 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
14144 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14145 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14146 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
14147 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14148 // CHECK19:       omp.inner.for.cond:
14149 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14150 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14151 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
14152 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14153 // CHECK19:       omp.inner.for.body:
14154 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14155 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
14156 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14157 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
14158 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14159 // CHECK19:       omp.body.continue:
14160 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14161 // CHECK19:       omp.inner.for.inc:
14162 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14163 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
14164 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
14165 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
14166 // CHECK19:       omp.inner.for.end:
14167 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
14168 // CHECK19:       omp.loop.exit:
14169 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
14170 // CHECK19-NEXT:    ret void
14171 //
14172 //
14173 // CHECK19-LABEL: define {{[^@]+}}@.omp_task_privates_map.
14174 // CHECK19-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
14175 // CHECK19-NEXT:  entry:
14176 // CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
14177 // CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
14178 // CHECK19-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
14179 // CHECK19-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
14180 // CHECK19-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
14181 // CHECK19-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
14182 // CHECK19-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
14183 // CHECK19-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
14184 // CHECK19-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
14185 // CHECK19-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
14186 // CHECK19-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
14187 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
14188 // CHECK19-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
14189 // CHECK19-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
14190 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
14191 // CHECK19-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
14192 // CHECK19-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
14193 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
14194 // CHECK19-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
14195 // CHECK19-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
14196 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
14197 // CHECK19-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
14198 // CHECK19-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
14199 // CHECK19-NEXT:    ret void
14200 //
14201 //
14202 // CHECK19-LABEL: define {{[^@]+}}@.omp_task_entry.
14203 // CHECK19-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
14204 // CHECK19-NEXT:  entry:
14205 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
14206 // CHECK19-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
14207 // CHECK19-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
14208 // CHECK19-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
14209 // CHECK19-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
14210 // CHECK19-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
14211 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
14212 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
14213 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
14214 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
14215 // CHECK19-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
14216 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
14217 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
14218 // CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
14219 // CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
14220 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
14221 // CHECK19-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
14222 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
14223 // CHECK19-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
14224 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
14225 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
14226 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
14227 // CHECK19-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
14228 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
14229 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
14230 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
14231 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
14232 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
14233 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
14234 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
14235 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
14236 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
14237 // CHECK19-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
14238 // CHECK19-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
14239 // CHECK19-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
14240 // CHECK19-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
14241 // CHECK19-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
14242 // CHECK19-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
14243 // CHECK19-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
14244 // CHECK19-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
14245 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
14246 // CHECK19-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
14247 // CHECK19-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !22
14248 // CHECK19-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !22
14249 // CHECK19-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !22
14250 // CHECK19-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !22
14251 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
14252 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
14253 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
14254 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
14255 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
14256 // CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
14257 // CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
14258 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
14259 // CHECK19-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
14260 // CHECK19-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
14261 // CHECK19-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
14262 // CHECK19:       omp_offload.failed.i:
14263 // CHECK19-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
14264 // CHECK19-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
14265 // CHECK19-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !22
14266 // CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !22
14267 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
14268 // CHECK19-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
14269 // CHECK19-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
14270 // CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
14271 // CHECK19-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
14272 // CHECK19-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
14273 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
14274 // CHECK19-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
14275 // CHECK19:       .omp_outlined..1.exit:
14276 // CHECK19-NEXT:    ret i32 0
14277 //
14278 //
14279 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
14280 // CHECK19-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
14281 // CHECK19-NEXT:  entry:
14282 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14283 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14284 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14285 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14286 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
14287 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
14288 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14289 // CHECK19-NEXT:    ret void
14290 //
14291 //
14292 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
14293 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
14294 // CHECK19-NEXT:  entry:
14295 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14296 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14297 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14298 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14299 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14300 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14301 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14302 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14303 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14304 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
14305 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14306 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14307 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14308 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14309 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
14310 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14311 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14312 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14313 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
14314 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14315 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14316 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
14317 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14318 // CHECK19:       cond.true:
14319 // CHECK19-NEXT:    br label [[COND_END:%.*]]
14320 // CHECK19:       cond.false:
14321 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14322 // CHECK19-NEXT:    br label [[COND_END]]
14323 // CHECK19:       cond.end:
14324 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
14325 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14326 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14327 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
14328 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14329 // CHECK19:       omp.inner.for.cond:
14330 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14331 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14332 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
14333 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14334 // CHECK19:       omp.inner.for.body:
14335 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14336 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
14337 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14338 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
14339 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14340 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
14341 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
14342 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14343 // CHECK19:       omp.body.continue:
14344 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14345 // CHECK19:       omp.inner.for.inc:
14346 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14347 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
14348 // CHECK19-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
14349 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
14350 // CHECK19:       omp.inner.for.end:
14351 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
14352 // CHECK19:       omp.loop.exit:
14353 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
14354 // CHECK19-NEXT:    ret void
14355 //
14356 //
14357 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
14358 // CHECK19-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
14359 // CHECK19-NEXT:  entry:
14360 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14361 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14362 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14363 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14364 // CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
14365 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14366 // CHECK19-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
14367 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14368 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14369 // CHECK19-NEXT:    ret void
14370 //
14371 //
14372 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
14373 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
14374 // CHECK19-NEXT:  entry:
14375 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14376 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14377 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14378 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14379 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14380 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14381 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14382 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14383 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14384 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
14385 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14386 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14387 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14388 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14389 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14390 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
14391 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14392 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14393 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14394 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
14395 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14396 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14397 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
14398 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14399 // CHECK19:       cond.true:
14400 // CHECK19-NEXT:    br label [[COND_END:%.*]]
14401 // CHECK19:       cond.false:
14402 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14403 // CHECK19-NEXT:    br label [[COND_END]]
14404 // CHECK19:       cond.end:
14405 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
14406 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14407 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14408 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
14409 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14410 // CHECK19:       omp.inner.for.cond:
14411 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14412 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14413 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
14414 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14415 // CHECK19:       omp.inner.for.body:
14416 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14417 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
14418 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14419 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
14420 // CHECK19-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
14421 // CHECK19-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
14422 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
14423 // CHECK19-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
14424 // CHECK19-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
14425 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14426 // CHECK19:       omp.body.continue:
14427 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14428 // CHECK19:       omp.inner.for.inc:
14429 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14430 // CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
14431 // CHECK19-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
14432 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
14433 // CHECK19:       omp.inner.for.end:
14434 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
14435 // CHECK19:       omp.loop.exit:
14436 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
14437 // CHECK19-NEXT:    ret void
14438 //
14439 //
14440 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
14441 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
14442 // CHECK19-NEXT:  entry:
14443 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14444 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14445 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14446 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14447 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14448 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14449 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14450 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14451 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
14452 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
14453 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
14454 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14455 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
14456 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14457 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
14458 // CHECK19-NEXT:    ret void
14459 //
14460 //
14461 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
14462 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
14463 // CHECK19-NEXT:  entry:
14464 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14465 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14466 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14467 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14468 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14469 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14470 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14471 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14472 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14473 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14474 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
14475 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14476 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14477 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14478 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14479 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14480 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14481 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
14482 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14483 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14484 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14485 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
14486 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14487 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14488 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
14489 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14490 // CHECK19:       cond.true:
14491 // CHECK19-NEXT:    br label [[COND_END:%.*]]
14492 // CHECK19:       cond.false:
14493 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14494 // CHECK19-NEXT:    br label [[COND_END]]
14495 // CHECK19:       cond.end:
14496 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
14497 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14498 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14499 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
14500 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14501 // CHECK19:       omp.inner.for.cond:
14502 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14503 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14504 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
14505 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14506 // CHECK19:       omp.inner.for.body:
14507 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14508 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
14509 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14510 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
14511 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14512 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
14513 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
14514 // CHECK19-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
14515 // CHECK19-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
14516 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
14517 // CHECK19-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
14518 // CHECK19-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
14519 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14520 // CHECK19:       omp.body.continue:
14521 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14522 // CHECK19:       omp.inner.for.inc:
14523 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14524 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
14525 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
14526 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
14527 // CHECK19:       omp.inner.for.end:
14528 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
14529 // CHECK19:       omp.loop.exit:
14530 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
14531 // CHECK19-NEXT:    ret void
14532 //
14533 //
14534 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
14535 // CHECK19-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14536 // CHECK19-NEXT:  entry:
14537 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14538 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14539 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14540 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14541 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14542 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14543 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14544 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14545 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14546 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14547 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14548 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
14549 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14550 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14551 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14552 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14553 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14554 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14555 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14556 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14557 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14558 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14559 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14560 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14561 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14562 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14563 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14564 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14565 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14566 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14567 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14568 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
14569 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
14570 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14571 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14572 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14573 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
14574 // CHECK19-NEXT:    ret void
14575 //
14576 //
14577 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
14578 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14579 // CHECK19-NEXT:  entry:
14580 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14581 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14582 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14583 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14584 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14585 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14586 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14587 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14588 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14589 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14590 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14591 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14592 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14593 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14594 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14595 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14596 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14597 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14598 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
14599 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14600 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14601 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14602 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14603 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14604 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14605 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14606 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14607 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14608 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14609 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14610 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14611 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14612 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14613 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14614 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14615 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14616 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14617 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14618 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14619 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14620 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
14621 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14622 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14623 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14624 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14625 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
14626 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
14627 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
14628 // CHECK19:       omp.dispatch.cond:
14629 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14630 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
14631 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14632 // CHECK19:       cond.true:
14633 // CHECK19-NEXT:    br label [[COND_END:%.*]]
14634 // CHECK19:       cond.false:
14635 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14636 // CHECK19-NEXT:    br label [[COND_END]]
14637 // CHECK19:       cond.end:
14638 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
14639 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14640 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14641 // CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
14642 // CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14643 // CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14644 // CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
14645 // CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
14646 // CHECK19:       omp.dispatch.body:
14647 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14648 // CHECK19:       omp.inner.for.cond:
14649 // CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
14650 // CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
14651 // CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
14652 // CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14653 // CHECK19:       omp.inner.for.body:
14654 // CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
14655 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
14656 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14657 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
14658 // CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !23
14659 // CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
14660 // CHECK19-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !23
14661 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
14662 // CHECK19-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
14663 // CHECK19-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
14664 // CHECK19-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
14665 // CHECK19-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
14666 // CHECK19-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
14667 // CHECK19-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
14668 // CHECK19-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
14669 // CHECK19-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
14670 // CHECK19-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
14671 // CHECK19-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
14672 // CHECK19-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
14673 // CHECK19-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
14674 // CHECK19-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
14675 // CHECK19-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
14676 // CHECK19-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
14677 // CHECK19-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
14678 // CHECK19-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
14679 // CHECK19-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
14680 // CHECK19-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
14681 // CHECK19-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
14682 // CHECK19-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
14683 // CHECK19-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
14684 // CHECK19-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
14685 // CHECK19-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !23
14686 // CHECK19-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
14687 // CHECK19-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !23
14688 // CHECK19-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
14689 // CHECK19-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !23
14690 // CHECK19-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
14691 // CHECK19-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
14692 // CHECK19-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
14693 // CHECK19-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !23
14694 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14695 // CHECK19:       omp.body.continue:
14696 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14697 // CHECK19:       omp.inner.for.inc:
14698 // CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
14699 // CHECK19-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
14700 // CHECK19-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
14701 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
14702 // CHECK19:       omp.inner.for.end:
14703 // CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
14704 // CHECK19:       omp.dispatch.inc:
14705 // CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14706 // CHECK19-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14707 // CHECK19-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
14708 // CHECK19-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
14709 // CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14710 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14711 // CHECK19-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
14712 // CHECK19-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
14713 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
14714 // CHECK19:       omp.dispatch.end:
14715 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
14716 // CHECK19-NEXT:    ret void
14717 //
14718 //
14719 // CHECK19-LABEL: define {{[^@]+}}@_Z3bari
14720 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
14721 // CHECK19-NEXT:  entry:
14722 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
14723 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
14724 // CHECK19-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
14725 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
14726 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
14727 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
14728 // CHECK19-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
14729 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
14730 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
14731 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
14732 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
14733 // CHECK19-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
14734 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
14735 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
14736 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
14737 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
14738 // CHECK19-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
14739 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
14740 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
14741 // CHECK19-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
14742 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
14743 // CHECK19-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
14744 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
14745 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
14746 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
14747 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
14748 // CHECK19-NEXT:    ret i32 [[TMP8]]
14749 //
14750 //
14751 // CHECK19-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
14752 // CHECK19-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
14753 // CHECK19-NEXT:  entry:
14754 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
14755 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
14756 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
14757 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
14758 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
14759 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
14760 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
14761 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
14762 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
14763 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
14764 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14765 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
14766 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
14767 // CHECK19-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
14768 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
14769 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
14770 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
14771 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
14772 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
14773 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
14774 // CHECK19-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
14775 // CHECK19-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
14776 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
14777 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
14778 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
14779 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
14780 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
14781 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
14782 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
14783 // CHECK19:       omp_if.then:
14784 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
14785 // CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
14786 // CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
14787 // CHECK19-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
14788 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
14789 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i32 40, i1 false)
14790 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14791 // CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
14792 // CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
14793 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14794 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
14795 // CHECK19-NEXT:    store double* [[A]], double** [[TMP14]], align 4
14796 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
14797 // CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
14798 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
14799 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
14800 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
14801 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
14802 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
14803 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
14804 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
14805 // CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
14806 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
14807 // CHECK19-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
14808 // CHECK19-NEXT:    store i32 2, i32* [[TMP22]], align 4
14809 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
14810 // CHECK19-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
14811 // CHECK19-NEXT:    store i32 2, i32* [[TMP24]], align 4
14812 // CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
14813 // CHECK19-NEXT:    store i8* null, i8** [[TMP25]], align 4
14814 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
14815 // CHECK19-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
14816 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
14817 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
14818 // CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
14819 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
14820 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
14821 // CHECK19-NEXT:    store i8* null, i8** [[TMP30]], align 4
14822 // CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
14823 // CHECK19-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
14824 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
14825 // CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
14826 // CHECK19-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
14827 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
14828 // CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
14829 // CHECK19-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
14830 // CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
14831 // CHECK19-NEXT:    store i8* null, i8** [[TMP36]], align 4
14832 // CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14833 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14834 // CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
14835 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
14836 // CHECK19-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14837 // CHECK19-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
14838 // CHECK19-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
14839 // CHECK19:       omp_offload.failed:
14840 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
14841 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
14842 // CHECK19:       omp_offload.cont:
14843 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
14844 // CHECK19:       omp_if.else:
14845 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
14846 // CHECK19-NEXT:    br label [[OMP_IF_END]]
14847 // CHECK19:       omp_if.end:
14848 // CHECK19-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
14849 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
14850 // CHECK19-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
14851 // CHECK19-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
14852 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
14853 // CHECK19-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
14854 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
14855 // CHECK19-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
14856 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
14857 // CHECK19-NEXT:    ret i32 [[ADD3]]
14858 //
14859 //
14860 // CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
14861 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
14862 // CHECK19-NEXT:  entry:
14863 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
14864 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
14865 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
14866 // CHECK19-NEXT:    [[AAA:%.*]] = alloca i8, align 1
14867 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
14868 // CHECK19-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
14869 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14870 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14871 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
14872 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
14873 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
14874 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
14875 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14876 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
14877 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
14878 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
14879 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
14880 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
14881 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
14882 // CHECK19-NEXT:    store i8 0, i8* [[AAA]], align 1
14883 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
14884 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[N_CASTED]], align 4
14885 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_CASTED]], align 4
14886 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
14887 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[A_CASTED]], align 4
14888 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
14889 // CHECK19-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
14890 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14891 // CHECK19-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
14892 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14893 // CHECK19-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
14894 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
14895 // CHECK19-NEXT:    store i8 [[TMP6]], i8* [[CONV1]], align 1
14896 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
14897 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
14898 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
14899 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
14900 // CHECK19:       omp_if.then:
14901 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14902 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
14903 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
14904 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14905 // CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
14906 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP12]], align 4
14907 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
14908 // CHECK19-NEXT:    store i8* null, i8** [[TMP13]], align 4
14909 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
14910 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
14911 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
14912 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
14913 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
14914 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP17]], align 4
14915 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
14916 // CHECK19-NEXT:    store i8* null, i8** [[TMP18]], align 4
14917 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
14918 // CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
14919 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
14920 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
14921 // CHECK19-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
14922 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
14923 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
14924 // CHECK19-NEXT:    store i8* null, i8** [[TMP23]], align 4
14925 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
14926 // CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
14927 // CHECK19-NEXT:    store i32 [[TMP7]], i32* [[TMP25]], align 4
14928 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
14929 // CHECK19-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
14930 // CHECK19-NEXT:    store i32 [[TMP7]], i32* [[TMP27]], align 4
14931 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
14932 // CHECK19-NEXT:    store i8* null, i8** [[TMP28]], align 4
14933 // CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
14934 // CHECK19-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
14935 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 4
14936 // CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
14937 // CHECK19-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
14938 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 4
14939 // CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
14940 // CHECK19-NEXT:    store i8* null, i8** [[TMP33]], align 4
14941 // CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14942 // CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14943 // CHECK19-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
14944 // CHECK19-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
14945 // CHECK19-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
14946 // CHECK19-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_2]], align 4
14947 // CHECK19-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
14948 // CHECK19-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
14949 // CHECK19-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
14950 // CHECK19-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
14951 // CHECK19-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
14952 // CHECK19-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
14953 // CHECK19-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
14954 // CHECK19-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
14955 // CHECK19-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
14956 // CHECK19-NEXT:    [[ADD6:%.*]] = add i32 [[TMP40]], 1
14957 // CHECK19-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD6]] to i64
14958 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
14959 // CHECK19-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14960 // CHECK19-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
14961 // CHECK19-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
14962 // CHECK19:       omp_offload.failed:
14963 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
14964 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
14965 // CHECK19:       omp_offload.cont:
14966 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
14967 // CHECK19:       omp_if.else:
14968 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
14969 // CHECK19-NEXT:    br label [[OMP_IF_END]]
14970 // CHECK19:       omp_if.end:
14971 // CHECK19-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
14972 // CHECK19-NEXT:    ret i32 [[TMP44]]
14973 //
14974 //
14975 // CHECK19-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
14976 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
14977 // CHECK19-NEXT:  entry:
14978 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
14979 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
14980 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
14981 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
14982 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14983 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14984 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
14985 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
14986 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
14987 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14988 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
14989 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
14990 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
14991 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
14992 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
14993 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
14994 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
14995 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14996 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
14997 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14998 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
14999 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
15000 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
15001 // CHECK19:       omp_if.then:
15002 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
15003 // CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
15004 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
15005 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
15006 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
15007 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
15008 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
15009 // CHECK19-NEXT:    store i8* null, i8** [[TMP9]], align 4
15010 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
15011 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
15012 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
15013 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
15014 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
15015 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
15016 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
15017 // CHECK19-NEXT:    store i8* null, i8** [[TMP14]], align 4
15018 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
15019 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
15020 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
15021 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
15022 // CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
15023 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
15024 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
15025 // CHECK19-NEXT:    store i8* null, i8** [[TMP19]], align 4
15026 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
15027 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
15028 // CHECK19-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
15029 // CHECK19-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15030 // CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
15031 // CHECK19-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
15032 // CHECK19:       omp_offload.failed:
15033 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
15034 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
15035 // CHECK19:       omp_offload.cont:
15036 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
15037 // CHECK19:       omp_if.else:
15038 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
15039 // CHECK19-NEXT:    br label [[OMP_IF_END]]
15040 // CHECK19:       omp_if.end:
15041 // CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
15042 // CHECK19-NEXT:    ret i32 [[TMP24]]
15043 //
15044 //
15045 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
15046 // CHECK19-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
15047 // CHECK19-NEXT:  entry:
15048 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
15049 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
15050 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15051 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
15052 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
15053 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
15054 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
15055 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
15056 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15057 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
15058 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
15059 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
15060 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15061 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
15062 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
15063 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
15064 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
15065 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
15066 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
15067 // CHECK19-NEXT:    ret void
15068 //
15069 //
15070 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..12
15071 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
15072 // CHECK19-NEXT:  entry:
15073 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15074 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15075 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
15076 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
15077 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15078 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
15079 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
15080 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
15081 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15082 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
15083 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
15084 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15085 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15086 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
15087 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15088 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15089 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
15090 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
15091 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15092 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
15093 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
15094 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
15095 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15096 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
15097 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
15098 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
15099 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
15100 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15101 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15102 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15103 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
15104 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15105 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15106 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
15107 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15108 // CHECK19:       cond.true:
15109 // CHECK19-NEXT:    br label [[COND_END:%.*]]
15110 // CHECK19:       cond.false:
15111 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15112 // CHECK19-NEXT:    br label [[COND_END]]
15113 // CHECK19:       cond.end:
15114 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
15115 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15116 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15117 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
15118 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
15119 // CHECK19:       omp.inner.for.cond:
15120 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15121 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15122 // CHECK19-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
15123 // CHECK19-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15124 // CHECK19:       omp.inner.for.body:
15125 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15126 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
15127 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15128 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
15129 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
15130 // CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
15131 // CHECK19-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
15132 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
15133 // CHECK19-NEXT:    store double [[ADD4]], double* [[A]], align 4
15134 // CHECK19-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
15135 // CHECK19-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
15136 // CHECK19-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
15137 // CHECK19-NEXT:    store double [[INC]], double* [[A5]], align 4
15138 // CHECK19-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
15139 // CHECK19-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
15140 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
15141 // CHECK19-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
15142 // CHECK19-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
15143 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
15144 // CHECK19:       omp.body.continue:
15145 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
15146 // CHECK19:       omp.inner.for.inc:
15147 // CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15148 // CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
15149 // CHECK19-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
15150 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
15151 // CHECK19:       omp.inner.for.end:
15152 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
15153 // CHECK19:       omp.loop.exit:
15154 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
15155 // CHECK19-NEXT:    ret void
15156 //
15157 //
15158 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
15159 // CHECK19-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
15160 // CHECK19-NEXT:  entry:
15161 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
15162 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15163 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15164 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
15165 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15166 // CHECK19-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
15167 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15168 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15169 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
15170 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
15171 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15172 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15173 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
15174 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15175 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15176 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
15177 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15178 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
15179 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
15180 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
15181 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
15182 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
15183 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
15184 // CHECK19-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
15185 // CHECK19-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15186 // CHECK19-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
15187 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15188 // CHECK19-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
15189 // CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
15190 // CHECK19-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
15191 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
15192 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
15193 // CHECK19-NEXT:    ret void
15194 //
15195 //
15196 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..15
15197 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
15198 // CHECK19-NEXT:  entry:
15199 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15200 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15201 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
15202 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15203 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15204 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
15205 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15206 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
15207 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15208 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15209 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
15210 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
15211 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
15212 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
15213 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
15214 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15215 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15216 // CHECK19-NEXT:    [[I6:%.*]] = alloca i32, align 4
15217 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15218 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15219 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
15220 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15221 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15222 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
15223 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15224 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15225 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
15226 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15227 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15228 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
15229 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
15230 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
15231 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15232 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15233 // CHECK19-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
15234 // CHECK19-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
15235 // CHECK19-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
15236 // CHECK19-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
15237 // CHECK19-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
15238 // CHECK19-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
15239 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15240 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
15241 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15242 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15243 // CHECK19-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
15244 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15245 // CHECK19:       omp.precond.then:
15246 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
15247 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
15248 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
15249 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15250 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15251 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15252 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
15253 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15254 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15255 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
15256 // CHECK19-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
15257 // CHECK19-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15258 // CHECK19:       cond.true:
15259 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
15260 // CHECK19-NEXT:    br label [[COND_END:%.*]]
15261 // CHECK19:       cond.false:
15262 // CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15263 // CHECK19-NEXT:    br label [[COND_END]]
15264 // CHECK19:       cond.end:
15265 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
15266 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15267 // CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15268 // CHECK19-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
15269 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
15270 // CHECK19:       omp.inner.for.cond:
15271 // CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15272 // CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15273 // CHECK19-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
15274 // CHECK19-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
15275 // CHECK19-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15276 // CHECK19:       omp.inner.for.body:
15277 // CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15278 // CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15279 // CHECK19-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
15280 // CHECK19-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
15281 // CHECK19-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
15282 // CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
15283 // CHECK19-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
15284 // CHECK19-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
15285 // CHECK19-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
15286 // CHECK19-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
15287 // CHECK19-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
15288 // CHECK19-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
15289 // CHECK19-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
15290 // CHECK19-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
15291 // CHECK19-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
15292 // CHECK19-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
15293 // CHECK19-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
15294 // CHECK19-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
15295 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
15296 // CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
15297 // CHECK19-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
15298 // CHECK19-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
15299 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
15300 // CHECK19:       omp.body.continue:
15301 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
15302 // CHECK19:       omp.inner.for.inc:
15303 // CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15304 // CHECK19-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
15305 // CHECK19-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
15306 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
15307 // CHECK19:       omp.inner.for.end:
15308 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
15309 // CHECK19:       omp.loop.exit:
15310 // CHECK19-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15311 // CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
15312 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
15313 // CHECK19-NEXT:    br label [[OMP_PRECOND_END]]
15314 // CHECK19:       omp.precond.end:
15315 // CHECK19-NEXT:    ret void
15316 //
15317 //
15318 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
15319 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
15320 // CHECK19-NEXT:  entry:
15321 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15322 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15323 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15324 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15325 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15326 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15327 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15328 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15329 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15330 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15331 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15332 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
15333 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
15334 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
15335 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15336 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
15337 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15338 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
15339 // CHECK19-NEXT:    ret void
15340 //
15341 //
15342 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..18
15343 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
15344 // CHECK19-NEXT:  entry:
15345 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15346 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15347 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15348 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15349 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15350 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
15351 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15352 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
15353 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
15354 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15355 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15356 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
15357 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15358 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15359 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15360 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15361 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15362 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15363 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15364 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
15365 // CHECK19-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
15366 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15367 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15368 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15369 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
15370 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15371 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15372 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
15373 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15374 // CHECK19:       cond.true:
15375 // CHECK19-NEXT:    br label [[COND_END:%.*]]
15376 // CHECK19:       cond.false:
15377 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15378 // CHECK19-NEXT:    br label [[COND_END]]
15379 // CHECK19:       cond.end:
15380 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
15381 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15382 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15383 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
15384 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
15385 // CHECK19:       omp.inner.for.cond:
15386 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15387 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15388 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
15389 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15390 // CHECK19:       omp.inner.for.body:
15391 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15392 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
15393 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15394 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
15395 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
15396 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
15397 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
15398 // CHECK19-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
15399 // CHECK19-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
15400 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
15401 // CHECK19-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
15402 // CHECK19-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
15403 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
15404 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
15405 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
15406 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
15407 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
15408 // CHECK19:       omp.body.continue:
15409 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
15410 // CHECK19:       omp.inner.for.inc:
15411 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15412 // CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
15413 // CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
15414 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]]
15415 // CHECK19:       omp.inner.for.end:
15416 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
15417 // CHECK19:       omp.loop.exit:
15418 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
15419 // CHECK19-NEXT:    ret void
15420 //
15421 //
15422 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
15423 // CHECK19-SAME: () #[[ATTR4]] {
15424 // CHECK19-NEXT:  entry:
15425 // CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
15426 // CHECK19-NEXT:    ret void
15427 //
15428 //
15429 // CHECK20-LABEL: define {{[^@]+}}@_Z3fooi
15430 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
15431 // CHECK20-NEXT:  entry:
15432 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
15433 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
15434 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
15435 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
15436 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
15437 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
15438 // CHECK20-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
15439 // CHECK20-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
15440 // CHECK20-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
15441 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15442 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
15443 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15444 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
15445 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
15446 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
15447 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
15448 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
15449 // CHECK20-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
15450 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15451 // CHECK20-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
15452 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
15453 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
15454 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
15455 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15456 // CHECK20-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
15457 // CHECK20-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
15458 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
15459 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
15460 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
15461 // CHECK20-NEXT:    [[_TMP15:%.*]] = alloca i32, align 4
15462 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_18:%.*]] = alloca i32, align 4
15463 // CHECK20-NEXT:    [[A_CASTED19:%.*]] = alloca i32, align 4
15464 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED20:%.*]] = alloca i32, align 4
15465 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS23:%.*]] = alloca [10 x i8*], align 4
15466 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS24:%.*]] = alloca [10 x i8*], align 4
15467 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS25:%.*]] = alloca [10 x i8*], align 4
15468 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
15469 // CHECK20-NEXT:    [[_TMP26:%.*]] = alloca i32, align 4
15470 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
15471 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
15472 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
15473 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
15474 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
15475 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
15476 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
15477 // CHECK20-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
15478 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
15479 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
15480 // CHECK20-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
15481 // CHECK20-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
15482 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
15483 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
15484 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
15485 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
15486 // CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
15487 // CHECK20-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
15488 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15489 // CHECK20-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
15490 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15491 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15492 // CHECK20-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15493 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15494 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15495 // CHECK20-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
15496 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
15497 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
15498 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
15499 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
15500 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
15501 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
15502 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
15503 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
15504 // CHECK20-NEXT:    store i8* null, i8** [[TMP17]], align 4
15505 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
15506 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
15507 // CHECK20-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
15508 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
15509 // CHECK20-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
15510 // CHECK20-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
15511 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
15512 // CHECK20-NEXT:    store i8* null, i8** [[TMP22]], align 4
15513 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
15514 // CHECK20-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
15515 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
15516 // CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
15517 // CHECK20-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
15518 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
15519 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
15520 // CHECK20-NEXT:    store i8* null, i8** [[TMP27]], align 4
15521 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
15522 // CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
15523 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
15524 // CHECK20-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
15525 // CHECK20-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
15526 // CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
15527 // CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15528 // CHECK20-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
15529 // CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
15530 // CHECK20-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15531 // CHECK20-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
15532 // CHECK20-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 72, i32 12, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
15533 // CHECK20-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
15534 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
15535 // CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
15536 // CHECK20-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
15537 // CHECK20-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
15538 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
15539 // CHECK20-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
15540 // CHECK20-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
15541 // CHECK20-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
15542 // CHECK20-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
15543 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
15544 // CHECK20-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
15545 // CHECK20-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
15546 // CHECK20-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
15547 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
15548 // CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
15549 // CHECK20-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
15550 // CHECK20-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
15551 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
15552 // CHECK20-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
15553 // CHECK20-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
15554 // CHECK20-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
15555 // CHECK20-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP36]])
15556 // CHECK20-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
15557 // CHECK20-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
15558 // CHECK20-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
15559 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
15560 // CHECK20-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
15561 // CHECK20-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
15562 // CHECK20-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
15563 // CHECK20-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
15564 // CHECK20-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
15565 // CHECK20-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
15566 // CHECK20-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
15567 // CHECK20-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
15568 // CHECK20-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
15569 // CHECK20-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
15570 // CHECK20-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
15571 // CHECK20-NEXT:    store i8* null, i8** [[TMP63]], align 4
15572 // CHECK20-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
15573 // CHECK20-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
15574 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
15575 // CHECK20-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113.region_id, i32 1, i8** [[TMP64]], i8** [[TMP65]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15576 // CHECK20-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
15577 // CHECK20-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
15578 // CHECK20:       omp_offload.failed:
15579 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113(i32 [[TMP58]]) #[[ATTR3]]
15580 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
15581 // CHECK20:       omp_offload.cont:
15582 // CHECK20-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
15583 // CHECK20-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
15584 // CHECK20-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
15585 // CHECK20-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
15586 // CHECK20-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
15587 // CHECK20-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
15588 // CHECK20-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
15589 // CHECK20-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
15590 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
15591 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
15592 // CHECK20:       omp_if.then:
15593 // CHECK20-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
15594 // CHECK20-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
15595 // CHECK20-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
15596 // CHECK20-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
15597 // CHECK20-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
15598 // CHECK20-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
15599 // CHECK20-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
15600 // CHECK20-NEXT:    store i8* null, i8** [[TMP77]], align 4
15601 // CHECK20-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
15602 // CHECK20-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
15603 // CHECK20-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
15604 // CHECK20-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
15605 // CHECK20-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
15606 // CHECK20-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
15607 // CHECK20-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
15608 // CHECK20-NEXT:    store i8* null, i8** [[TMP82]], align 4
15609 // CHECK20-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
15610 // CHECK20-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
15611 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
15612 // CHECK20-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120.region_id, i32 2, i8** [[TMP83]], i8** [[TMP84]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.7, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.8, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15613 // CHECK20-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
15614 // CHECK20-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
15615 // CHECK20:       omp_offload.failed16:
15616 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
15617 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT17]]
15618 // CHECK20:       omp_offload.cont17:
15619 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
15620 // CHECK20:       omp_if.else:
15621 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
15622 // CHECK20-NEXT:    br label [[OMP_IF_END]]
15623 // CHECK20:       omp_if.end:
15624 // CHECK20-NEXT:    [[TMP87:%.*]] = load i32, i32* [[N_ADDR]], align 4
15625 // CHECK20-NEXT:    store i32 [[TMP87]], i32* [[DOTCAPTURE_EXPR_18]], align 4
15626 // CHECK20-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A]], align 4
15627 // CHECK20-NEXT:    store i32 [[TMP88]], i32* [[A_CASTED19]], align 4
15628 // CHECK20-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A_CASTED19]], align 4
15629 // CHECK20-NEXT:    [[TMP90:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_18]], align 4
15630 // CHECK20-NEXT:    store i32 [[TMP90]], i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
15631 // CHECK20-NEXT:    [[TMP91:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED20]], align 4
15632 // CHECK20-NEXT:    [[TMP92:%.*]] = load i32, i32* [[N_ADDR]], align 4
15633 // CHECK20-NEXT:    [[CMP21:%.*]] = icmp sgt i32 [[TMP92]], 20
15634 // CHECK20-NEXT:    br i1 [[CMP21]], label [[OMP_IF_THEN22:%.*]], label [[OMP_IF_ELSE29:%.*]]
15635 // CHECK20:       omp_if.then22:
15636 // CHECK20-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP1]], 4
15637 // CHECK20-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
15638 // CHECK20-NEXT:    [[TMP95:%.*]] = mul nuw i32 5, [[TMP3]]
15639 // CHECK20-NEXT:    [[TMP96:%.*]] = mul nuw i32 [[TMP95]], 8
15640 // CHECK20-NEXT:    [[TMP97:%.*]] = sext i32 [[TMP96]] to i64
15641 // CHECK20-NEXT:    [[TMP98:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
15642 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP98]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.10 to i8*), i32 80, i1 false)
15643 // CHECK20-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
15644 // CHECK20-NEXT:    [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
15645 // CHECK20-NEXT:    store i32 [[TMP89]], i32* [[TMP100]], align 4
15646 // CHECK20-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
15647 // CHECK20-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to i32*
15648 // CHECK20-NEXT:    store i32 [[TMP89]], i32* [[TMP102]], align 4
15649 // CHECK20-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 0
15650 // CHECK20-NEXT:    store i8* null, i8** [[TMP103]], align 4
15651 // CHECK20-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 1
15652 // CHECK20-NEXT:    [[TMP105:%.*]] = bitcast i8** [[TMP104]] to [10 x float]**
15653 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP105]], align 4
15654 // CHECK20-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 1
15655 // CHECK20-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to [10 x float]**
15656 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP107]], align 4
15657 // CHECK20-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 1
15658 // CHECK20-NEXT:    store i8* null, i8** [[TMP108]], align 4
15659 // CHECK20-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 2
15660 // CHECK20-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
15661 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP110]], align 4
15662 // CHECK20-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 2
15663 // CHECK20-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to i32*
15664 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP112]], align 4
15665 // CHECK20-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 2
15666 // CHECK20-NEXT:    store i8* null, i8** [[TMP113]], align 4
15667 // CHECK20-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 3
15668 // CHECK20-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to float**
15669 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP115]], align 4
15670 // CHECK20-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 3
15671 // CHECK20-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to float**
15672 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP117]], align 4
15673 // CHECK20-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
15674 // CHECK20-NEXT:    store i64 [[TMP94]], i64* [[TMP118]], align 4
15675 // CHECK20-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 3
15676 // CHECK20-NEXT:    store i8* null, i8** [[TMP119]], align 4
15677 // CHECK20-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 4
15678 // CHECK20-NEXT:    [[TMP121:%.*]] = bitcast i8** [[TMP120]] to [5 x [10 x double]]**
15679 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP121]], align 4
15680 // CHECK20-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 4
15681 // CHECK20-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to [5 x [10 x double]]**
15682 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP123]], align 4
15683 // CHECK20-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 4
15684 // CHECK20-NEXT:    store i8* null, i8** [[TMP124]], align 4
15685 // CHECK20-NEXT:    [[TMP125:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 5
15686 // CHECK20-NEXT:    [[TMP126:%.*]] = bitcast i8** [[TMP125]] to i32*
15687 // CHECK20-NEXT:    store i32 5, i32* [[TMP126]], align 4
15688 // CHECK20-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 5
15689 // CHECK20-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
15690 // CHECK20-NEXT:    store i32 5, i32* [[TMP128]], align 4
15691 // CHECK20-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 5
15692 // CHECK20-NEXT:    store i8* null, i8** [[TMP129]], align 4
15693 // CHECK20-NEXT:    [[TMP130:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 6
15694 // CHECK20-NEXT:    [[TMP131:%.*]] = bitcast i8** [[TMP130]] to i32*
15695 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP131]], align 4
15696 // CHECK20-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 6
15697 // CHECK20-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to i32*
15698 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP133]], align 4
15699 // CHECK20-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 6
15700 // CHECK20-NEXT:    store i8* null, i8** [[TMP134]], align 4
15701 // CHECK20-NEXT:    [[TMP135:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 7
15702 // CHECK20-NEXT:    [[TMP136:%.*]] = bitcast i8** [[TMP135]] to double**
15703 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP136]], align 4
15704 // CHECK20-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 7
15705 // CHECK20-NEXT:    [[TMP138:%.*]] = bitcast i8** [[TMP137]] to double**
15706 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP138]], align 4
15707 // CHECK20-NEXT:    [[TMP139:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
15708 // CHECK20-NEXT:    store i64 [[TMP97]], i64* [[TMP139]], align 4
15709 // CHECK20-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 7
15710 // CHECK20-NEXT:    store i8* null, i8** [[TMP140]], align 4
15711 // CHECK20-NEXT:    [[TMP141:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 8
15712 // CHECK20-NEXT:    [[TMP142:%.*]] = bitcast i8** [[TMP141]] to %struct.TT**
15713 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP142]], align 4
15714 // CHECK20-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 8
15715 // CHECK20-NEXT:    [[TMP144:%.*]] = bitcast i8** [[TMP143]] to %struct.TT**
15716 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP144]], align 4
15717 // CHECK20-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 8
15718 // CHECK20-NEXT:    store i8* null, i8** [[TMP145]], align 4
15719 // CHECK20-NEXT:    [[TMP146:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 9
15720 // CHECK20-NEXT:    [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
15721 // CHECK20-NEXT:    store i32 [[TMP91]], i32* [[TMP147]], align 4
15722 // CHECK20-NEXT:    [[TMP148:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 9
15723 // CHECK20-NEXT:    [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
15724 // CHECK20-NEXT:    store i32 [[TMP91]], i32* [[TMP149]], align 4
15725 // CHECK20-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS25]], i32 0, i32 9
15726 // CHECK20-NEXT:    store i8* null, i8** [[TMP150]], align 4
15727 // CHECK20-NEXT:    [[TMP151:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS23]], i32 0, i32 0
15728 // CHECK20-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS24]], i32 0, i32 0
15729 // CHECK20-NEXT:    [[TMP153:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
15730 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
15731 // CHECK20-NEXT:    [[TMP154:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145.region_id, i32 10, i8** [[TMP151]], i8** [[TMP152]], i64* [[TMP153]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15732 // CHECK20-NEXT:    [[TMP155:%.*]] = icmp ne i32 [[TMP154]], 0
15733 // CHECK20-NEXT:    br i1 [[TMP155]], label [[OMP_OFFLOAD_FAILED27:%.*]], label [[OMP_OFFLOAD_CONT28:%.*]]
15734 // CHECK20:       omp_offload.failed27:
15735 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
15736 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT28]]
15737 // CHECK20:       omp_offload.cont28:
15738 // CHECK20-NEXT:    br label [[OMP_IF_END30:%.*]]
15739 // CHECK20:       omp_if.else29:
15740 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145(i32 [[TMP89]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP91]]) #[[ATTR3]]
15741 // CHECK20-NEXT:    br label [[OMP_IF_END30]]
15742 // CHECK20:       omp_if.end30:
15743 // CHECK20-NEXT:    [[TMP156:%.*]] = load i32, i32* [[A]], align 4
15744 // CHECK20-NEXT:    [[TMP157:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
15745 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP157]])
15746 // CHECK20-NEXT:    ret i32 [[TMP156]]
15747 //
15748 //
15749 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
15750 // CHECK20-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
15751 // CHECK20-NEXT:  entry:
15752 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15753 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
15754 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
15755 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15756 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
15757 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15758 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15759 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
15760 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15761 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15762 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
15763 // CHECK20-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
15764 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
15765 // CHECK20-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15766 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
15767 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15768 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
15769 // CHECK20-NEXT:    ret void
15770 //
15771 //
15772 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
15773 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
15774 // CHECK20-NEXT:  entry:
15775 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15776 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15777 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15778 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
15779 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15780 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
15781 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
15782 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15783 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15784 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
15785 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15786 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15787 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15788 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15789 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
15790 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
15791 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15792 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15793 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15794 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
15795 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15796 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15797 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
15798 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15799 // CHECK20:       cond.true:
15800 // CHECK20-NEXT:    br label [[COND_END:%.*]]
15801 // CHECK20:       cond.false:
15802 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15803 // CHECK20-NEXT:    br label [[COND_END]]
15804 // CHECK20:       cond.end:
15805 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
15806 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15807 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15808 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
15809 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
15810 // CHECK20:       omp.inner.for.cond:
15811 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15812 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15813 // CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
15814 // CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15815 // CHECK20:       omp.inner.for.body:
15816 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15817 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
15818 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15819 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
15820 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
15821 // CHECK20:       omp.body.continue:
15822 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
15823 // CHECK20:       omp.inner.for.inc:
15824 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15825 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
15826 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
15827 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
15828 // CHECK20:       omp.inner.for.end:
15829 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
15830 // CHECK20:       omp.loop.exit:
15831 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
15832 // CHECK20-NEXT:    ret void
15833 //
15834 //
15835 // CHECK20-LABEL: define {{[^@]+}}@.omp_task_privates_map.
15836 // CHECK20-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i16** noalias noundef [[TMP1:%.*]], [3 x i8*]** noalias noundef [[TMP2:%.*]], [3 x i8*]** noalias noundef [[TMP3:%.*]], [3 x i64]** noalias noundef [[TMP4:%.*]]) #[[ATTR4:[0-9]+]] {
15837 // CHECK20-NEXT:  entry:
15838 // CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
15839 // CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
15840 // CHECK20-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
15841 // CHECK20-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
15842 // CHECK20-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
15843 // CHECK20-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
15844 // CHECK20-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
15845 // CHECK20-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
15846 // CHECK20-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
15847 // CHECK20-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
15848 // CHECK20-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
15849 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
15850 // CHECK20-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
15851 // CHECK20-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
15852 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
15853 // CHECK20-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
15854 // CHECK20-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
15855 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
15856 // CHECK20-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
15857 // CHECK20-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
15858 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
15859 // CHECK20-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
15860 // CHECK20-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
15861 // CHECK20-NEXT:    ret void
15862 //
15863 //
15864 // CHECK20-LABEL: define {{[^@]+}}@.omp_task_entry.
15865 // CHECK20-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
15866 // CHECK20-NEXT:  entry:
15867 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
15868 // CHECK20-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
15869 // CHECK20-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
15870 // CHECK20-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
15871 // CHECK20-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
15872 // CHECK20-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
15873 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
15874 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
15875 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
15876 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
15877 // CHECK20-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
15878 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
15879 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
15880 // CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
15881 // CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
15882 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
15883 // CHECK20-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
15884 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
15885 // CHECK20-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
15886 // CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
15887 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
15888 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
15889 // CHECK20-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
15890 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
15891 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
15892 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
15893 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
15894 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META13:![0-9]+]])
15895 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
15896 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
15897 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
15898 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !22
15899 // CHECK20-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !22
15900 // CHECK20-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
15901 // CHECK20-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
15902 // CHECK20-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !22
15903 // CHECK20-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
15904 // CHECK20-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !22
15905 // CHECK20-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !22
15906 // CHECK20-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !22
15907 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
15908 // CHECK20-NEXT:    call void [[TMP15]](i8* [[TMP14]], i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]]) #[[ATTR3]]
15909 // CHECK20-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !22
15910 // CHECK20-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !22
15911 // CHECK20-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !22
15912 // CHECK20-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !22
15913 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
15914 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
15915 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
15916 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
15917 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
15918 // CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
15919 // CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
15920 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10) #[[ATTR3]]
15921 // CHECK20-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* [[TMP22]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP25]], i32 [[TMP26]], i32 0, i8* null, i32 0, i8* null) #[[ATTR3]]
15922 // CHECK20-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
15923 // CHECK20-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
15924 // CHECK20:       omp_offload.failed.i:
15925 // CHECK20-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
15926 // CHECK20-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
15927 // CHECK20-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !22
15928 // CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !22
15929 // CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
15930 // CHECK20-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
15931 // CHECK20-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !22
15932 // CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
15933 // CHECK20-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
15934 // CHECK20-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !22
15935 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
15936 // CHECK20-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
15937 // CHECK20:       .omp_outlined..1.exit:
15938 // CHECK20-NEXT:    ret i32 0
15939 //
15940 //
15941 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l107
15942 // CHECK20-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
15943 // CHECK20-NEXT:  entry:
15944 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15945 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15946 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15947 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
15948 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
15949 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
15950 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
15951 // CHECK20-NEXT:    ret void
15952 //
15953 //
15954 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
15955 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
15956 // CHECK20-NEXT:  entry:
15957 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15958 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15959 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15960 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
15961 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
15962 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
15963 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
15964 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15965 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15966 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
15967 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15968 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15969 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15970 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
15971 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
15972 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15973 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15974 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15975 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
15976 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15977 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15978 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
15979 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15980 // CHECK20:       cond.true:
15981 // CHECK20-NEXT:    br label [[COND_END:%.*]]
15982 // CHECK20:       cond.false:
15983 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15984 // CHECK20-NEXT:    br label [[COND_END]]
15985 // CHECK20:       cond.end:
15986 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
15987 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15988 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15989 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
15990 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
15991 // CHECK20:       omp.inner.for.cond:
15992 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15993 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15994 // CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
15995 // CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15996 // CHECK20:       omp.inner.for.body:
15997 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15998 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
15999 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16000 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
16001 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
16002 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
16003 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
16004 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16005 // CHECK20:       omp.body.continue:
16006 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16007 // CHECK20:       omp.inner.for.inc:
16008 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16009 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
16010 // CHECK20-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
16011 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
16012 // CHECK20:       omp.inner.for.end:
16013 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
16014 // CHECK20:       omp.loop.exit:
16015 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
16016 // CHECK20-NEXT:    ret void
16017 //
16018 //
16019 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
16020 // CHECK20-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
16021 // CHECK20-NEXT:  entry:
16022 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16023 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16024 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16025 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16026 // CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
16027 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16028 // CHECK20-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
16029 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
16030 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
16031 // CHECK20-NEXT:    ret void
16032 //
16033 //
16034 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
16035 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
16036 // CHECK20-NEXT:  entry:
16037 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16038 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16039 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16040 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
16041 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16042 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
16043 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
16044 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16045 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16046 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
16047 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16048 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16049 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16050 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16051 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
16052 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
16053 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16054 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16055 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16056 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
16057 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16058 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16059 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
16060 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16061 // CHECK20:       cond.true:
16062 // CHECK20-NEXT:    br label [[COND_END:%.*]]
16063 // CHECK20:       cond.false:
16064 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16065 // CHECK20-NEXT:    br label [[COND_END]]
16066 // CHECK20:       cond.end:
16067 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
16068 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16069 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16070 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
16071 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
16072 // CHECK20:       omp.inner.for.cond:
16073 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16074 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16075 // CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
16076 // CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16077 // CHECK20:       omp.inner.for.body:
16078 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16079 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
16080 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16081 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
16082 // CHECK20-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
16083 // CHECK20-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
16084 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
16085 // CHECK20-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
16086 // CHECK20-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
16087 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16088 // CHECK20:       omp.body.continue:
16089 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16090 // CHECK20:       omp.inner.for.inc:
16091 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16092 // CHECK20-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
16093 // CHECK20-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
16094 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
16095 // CHECK20:       omp.inner.for.end:
16096 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
16097 // CHECK20:       omp.loop.exit:
16098 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
16099 // CHECK20-NEXT:    ret void
16100 //
16101 //
16102 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
16103 // CHECK20-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
16104 // CHECK20-NEXT:  entry:
16105 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16106 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16107 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16108 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16109 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16110 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16111 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16112 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
16113 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
16114 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
16115 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
16116 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16117 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
16118 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
16119 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
16120 // CHECK20-NEXT:    ret void
16121 //
16122 //
16123 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..6
16124 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
16125 // CHECK20-NEXT:  entry:
16126 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16127 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16128 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16129 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16130 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
16131 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16132 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
16133 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
16134 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16135 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16136 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
16137 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16138 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16139 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16140 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16141 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16142 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
16143 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
16144 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16145 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16146 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16147 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
16148 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16149 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16150 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
16151 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16152 // CHECK20:       cond.true:
16153 // CHECK20-NEXT:    br label [[COND_END:%.*]]
16154 // CHECK20:       cond.false:
16155 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16156 // CHECK20-NEXT:    br label [[COND_END]]
16157 // CHECK20:       cond.end:
16158 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
16159 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16160 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16161 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
16162 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
16163 // CHECK20:       omp.inner.for.cond:
16164 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16165 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16166 // CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
16167 // CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16168 // CHECK20:       omp.inner.for.body:
16169 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16170 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
16171 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16172 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
16173 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
16174 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
16175 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
16176 // CHECK20-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
16177 // CHECK20-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
16178 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
16179 // CHECK20-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
16180 // CHECK20-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
16181 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16182 // CHECK20:       omp.body.continue:
16183 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16184 // CHECK20:       omp.inner.for.inc:
16185 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16186 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
16187 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
16188 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
16189 // CHECK20:       omp.inner.for.end:
16190 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
16191 // CHECK20:       omp.loop.exit:
16192 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
16193 // CHECK20-NEXT:    ret void
16194 //
16195 //
16196 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
16197 // CHECK20-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16198 // CHECK20-NEXT:  entry:
16199 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16200 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
16201 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
16202 // CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
16203 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
16204 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
16205 // CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
16206 // CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
16207 // CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
16208 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16209 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16210 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16211 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16212 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
16213 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16214 // CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
16215 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
16216 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
16217 // CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
16218 // CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
16219 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
16220 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16221 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
16222 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16223 // CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
16224 // CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
16225 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
16226 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
16227 // CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
16228 // CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
16229 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
16230 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
16231 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
16232 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16233 // CHECK20-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16234 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16235 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
16236 // CHECK20-NEXT:    ret void
16237 //
16238 //
16239 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
16240 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16241 // CHECK20-NEXT:  entry:
16242 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16243 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16244 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16245 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
16246 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
16247 // CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
16248 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
16249 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
16250 // CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
16251 // CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
16252 // CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
16253 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16254 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
16255 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16256 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
16257 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
16258 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16259 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16260 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
16261 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16262 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16263 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16264 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
16265 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16266 // CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
16267 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
16268 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
16269 // CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
16270 // CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
16271 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
16272 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16273 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
16274 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16275 // CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
16276 // CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
16277 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
16278 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
16279 // CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
16280 // CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
16281 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
16282 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
16283 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16284 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16285 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16286 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16287 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
16288 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
16289 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
16290 // CHECK20:       omp.dispatch.cond:
16291 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16292 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
16293 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16294 // CHECK20:       cond.true:
16295 // CHECK20-NEXT:    br label [[COND_END:%.*]]
16296 // CHECK20:       cond.false:
16297 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16298 // CHECK20-NEXT:    br label [[COND_END]]
16299 // CHECK20:       cond.end:
16300 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
16301 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16302 // CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16303 // CHECK20-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
16304 // CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16305 // CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16306 // CHECK20-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
16307 // CHECK20-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
16308 // CHECK20:       omp.dispatch.body:
16309 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
16310 // CHECK20:       omp.inner.for.cond:
16311 // CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
16312 // CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
16313 // CHECK20-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
16314 // CHECK20-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16315 // CHECK20:       omp.inner.for.body:
16316 // CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
16317 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
16318 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16319 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
16320 // CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !23
16321 // CHECK20-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
16322 // CHECK20-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !23
16323 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
16324 // CHECK20-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
16325 // CHECK20-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
16326 // CHECK20-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
16327 // CHECK20-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
16328 // CHECK20-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
16329 // CHECK20-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
16330 // CHECK20-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
16331 // CHECK20-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
16332 // CHECK20-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
16333 // CHECK20-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
16334 // CHECK20-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !23
16335 // CHECK20-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
16336 // CHECK20-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
16337 // CHECK20-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
16338 // CHECK20-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
16339 // CHECK20-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !23
16340 // CHECK20-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
16341 // CHECK20-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
16342 // CHECK20-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
16343 // CHECK20-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
16344 // CHECK20-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
16345 // CHECK20-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !23
16346 // CHECK20-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
16347 // CHECK20-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !23
16348 // CHECK20-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
16349 // CHECK20-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !23
16350 // CHECK20-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
16351 // CHECK20-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !23
16352 // CHECK20-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
16353 // CHECK20-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
16354 // CHECK20-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
16355 // CHECK20-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !23
16356 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16357 // CHECK20:       omp.body.continue:
16358 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16359 // CHECK20:       omp.inner.for.inc:
16360 // CHECK20-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
16361 // CHECK20-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
16362 // CHECK20-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
16363 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
16364 // CHECK20:       omp.inner.for.end:
16365 // CHECK20-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
16366 // CHECK20:       omp.dispatch.inc:
16367 // CHECK20-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16368 // CHECK20-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16369 // CHECK20-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
16370 // CHECK20-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
16371 // CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16372 // CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16373 // CHECK20-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
16374 // CHECK20-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
16375 // CHECK20-NEXT:    br label [[OMP_DISPATCH_COND]]
16376 // CHECK20:       omp.dispatch.end:
16377 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
16378 // CHECK20-NEXT:    ret void
16379 //
16380 //
16381 // CHECK20-LABEL: define {{[^@]+}}@_Z3bari
16382 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
16383 // CHECK20-NEXT:  entry:
16384 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16385 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
16386 // CHECK20-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
16387 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16388 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
16389 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
16390 // CHECK20-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
16391 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
16392 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
16393 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
16394 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
16395 // CHECK20-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
16396 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
16397 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
16398 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
16399 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
16400 // CHECK20-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
16401 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
16402 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
16403 // CHECK20-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
16404 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
16405 // CHECK20-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
16406 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
16407 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
16408 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
16409 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
16410 // CHECK20-NEXT:    ret i32 [[TMP8]]
16411 //
16412 //
16413 // CHECK20-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
16414 // CHECK20-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
16415 // CHECK20-NEXT:  entry:
16416 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
16417 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16418 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
16419 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
16420 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
16421 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
16422 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
16423 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
16424 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
16425 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
16426 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16427 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
16428 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16429 // CHECK20-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
16430 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
16431 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
16432 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
16433 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
16434 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
16435 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
16436 // CHECK20-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
16437 // CHECK20-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
16438 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
16439 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
16440 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
16441 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
16442 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
16443 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
16444 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
16445 // CHECK20:       omp_if.then:
16446 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
16447 // CHECK20-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
16448 // CHECK20-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
16449 // CHECK20-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
16450 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
16451 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.13 to i8*), i32 40, i1 false)
16452 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16453 // CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
16454 // CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
16455 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16456 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
16457 // CHECK20-NEXT:    store double* [[A]], double** [[TMP14]], align 4
16458 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
16459 // CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
16460 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
16461 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
16462 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
16463 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
16464 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
16465 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
16466 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
16467 // CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
16468 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
16469 // CHECK20-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
16470 // CHECK20-NEXT:    store i32 2, i32* [[TMP22]], align 4
16471 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
16472 // CHECK20-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
16473 // CHECK20-NEXT:    store i32 2, i32* [[TMP24]], align 4
16474 // CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
16475 // CHECK20-NEXT:    store i8* null, i8** [[TMP25]], align 4
16476 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
16477 // CHECK20-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
16478 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
16479 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
16480 // CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
16481 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
16482 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
16483 // CHECK20-NEXT:    store i8* null, i8** [[TMP30]], align 4
16484 // CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
16485 // CHECK20-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
16486 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
16487 // CHECK20-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
16488 // CHECK20-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
16489 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
16490 // CHECK20-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
16491 // CHECK20-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
16492 // CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
16493 // CHECK20-NEXT:    store i8* null, i8** [[TMP36]], align 4
16494 // CHECK20-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16495 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16496 // CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
16497 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
16498 // CHECK20-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16499 // CHECK20-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
16500 // CHECK20-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
16501 // CHECK20:       omp_offload.failed:
16502 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
16503 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
16504 // CHECK20:       omp_offload.cont:
16505 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
16506 // CHECK20:       omp_if.else:
16507 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
16508 // CHECK20-NEXT:    br label [[OMP_IF_END]]
16509 // CHECK20:       omp_if.end:
16510 // CHECK20-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
16511 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
16512 // CHECK20-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
16513 // CHECK20-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
16514 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
16515 // CHECK20-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
16516 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
16517 // CHECK20-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
16518 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
16519 // CHECK20-NEXT:    ret i32 [[ADD3]]
16520 //
16521 //
16522 // CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
16523 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
16524 // CHECK20-NEXT:  entry:
16525 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16526 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
16527 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
16528 // CHECK20-NEXT:    [[AAA:%.*]] = alloca i8, align 1
16529 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
16530 // CHECK20-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
16531 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16532 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16533 // CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
16534 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
16535 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
16536 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
16537 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16538 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
16539 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
16540 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
16541 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16542 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
16543 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
16544 // CHECK20-NEXT:    store i8 0, i8* [[AAA]], align 1
16545 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
16546 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[N_CASTED]], align 4
16547 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_CASTED]], align 4
16548 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A]], align 4
16549 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[A_CASTED]], align 4
16550 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_CASTED]], align 4
16551 // CHECK20-NEXT:    [[TMP4:%.*]] = load i16, i16* [[AA]], align 2
16552 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16553 // CHECK20-NEXT:    store i16 [[TMP4]], i16* [[CONV]], align 2
16554 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AA_CASTED]], align 4
16555 // CHECK20-NEXT:    [[TMP6:%.*]] = load i8, i8* [[AAA]], align 1
16556 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
16557 // CHECK20-NEXT:    store i8 [[TMP6]], i8* [[CONV1]], align 1
16558 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
16559 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[N_ADDR]], align 4
16560 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP8]], 50
16561 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
16562 // CHECK20:       omp_if.then:
16563 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16564 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
16565 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
16566 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16567 // CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
16568 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP12]], align 4
16569 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
16570 // CHECK20-NEXT:    store i8* null, i8** [[TMP13]], align 4
16571 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
16572 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
16573 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
16574 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
16575 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
16576 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP17]], align 4
16577 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
16578 // CHECK20-NEXT:    store i8* null, i8** [[TMP18]], align 4
16579 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
16580 // CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
16581 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
16582 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
16583 // CHECK20-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
16584 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP22]], align 4
16585 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
16586 // CHECK20-NEXT:    store i8* null, i8** [[TMP23]], align 4
16587 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
16588 // CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
16589 // CHECK20-NEXT:    store i32 [[TMP7]], i32* [[TMP25]], align 4
16590 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
16591 // CHECK20-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
16592 // CHECK20-NEXT:    store i32 [[TMP7]], i32* [[TMP27]], align 4
16593 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
16594 // CHECK20-NEXT:    store i8* null, i8** [[TMP28]], align 4
16595 // CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
16596 // CHECK20-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [10 x i32]**
16597 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP30]], align 4
16598 // CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
16599 // CHECK20-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to [10 x i32]**
16600 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP32]], align 4
16601 // CHECK20-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
16602 // CHECK20-NEXT:    store i8* null, i8** [[TMP33]], align 4
16603 // CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16604 // CHECK20-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16605 // CHECK20-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
16606 // CHECK20-NEXT:    store i32 [[TMP36]], i32* [[DOTCAPTURE_EXPR_]], align 4
16607 // CHECK20-NEXT:    [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
16608 // CHECK20-NEXT:    store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_2]], align 4
16609 // CHECK20-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16610 // CHECK20-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16611 // CHECK20-NEXT:    [[SUB:%.*]] = sub i32 [[TMP38]], [[TMP39]]
16612 // CHECK20-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
16613 // CHECK20-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
16614 // CHECK20-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
16615 // CHECK20-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
16616 // CHECK20-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
16617 // CHECK20-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
16618 // CHECK20-NEXT:    [[ADD6:%.*]] = add i32 [[TMP40]], 1
16619 // CHECK20-NEXT:    [[TMP41:%.*]] = zext i32 [[ADD6]] to i64
16620 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP41]])
16621 // CHECK20-NEXT:    [[TMP42:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200.region_id, i32 5, i8** [[TMP34]], i8** [[TMP35]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16622 // CHECK20-NEXT:    [[TMP43:%.*]] = icmp ne i32 [[TMP42]], 0
16623 // CHECK20-NEXT:    br i1 [[TMP43]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
16624 // CHECK20:       omp_offload.failed:
16625 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
16626 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
16627 // CHECK20:       omp_offload.cont:
16628 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
16629 // CHECK20:       omp_if.else:
16630 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP7]], [10 x i32]* [[B]]) #[[ATTR3]]
16631 // CHECK20-NEXT:    br label [[OMP_IF_END]]
16632 // CHECK20:       omp_if.end:
16633 // CHECK20-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
16634 // CHECK20-NEXT:    ret i32 [[TMP44]]
16635 //
16636 //
16637 // CHECK20-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
16638 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
16639 // CHECK20-NEXT:  entry:
16640 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16641 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
16642 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
16643 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
16644 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16645 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16646 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
16647 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
16648 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
16649 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16650 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16651 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
16652 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
16653 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
16654 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
16655 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
16656 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
16657 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16658 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
16659 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
16660 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
16661 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
16662 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
16663 // CHECK20:       omp_if.then:
16664 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16665 // CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
16666 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
16667 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16668 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
16669 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
16670 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
16671 // CHECK20-NEXT:    store i8* null, i8** [[TMP9]], align 4
16672 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
16673 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
16674 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
16675 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
16676 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
16677 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
16678 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
16679 // CHECK20-NEXT:    store i8* null, i8** [[TMP14]], align 4
16680 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
16681 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
16682 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
16683 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
16684 // CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
16685 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
16686 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
16687 // CHECK20-NEXT:    store i8* null, i8** [[TMP19]], align 4
16688 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16689 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16690 // CHECK20-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
16691 // CHECK20-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16692 // CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
16693 // CHECK20-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
16694 // CHECK20:       omp_offload.failed:
16695 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
16696 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
16697 // CHECK20:       omp_offload.cont:
16698 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
16699 // CHECK20:       omp_if.else:
16700 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
16701 // CHECK20-NEXT:    br label [[OMP_IF_END]]
16702 // CHECK20:       omp_if.end:
16703 // CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
16704 // CHECK20-NEXT:    ret i32 [[TMP24]]
16705 //
16706 //
16707 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
16708 // CHECK20-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
16709 // CHECK20-NEXT:  entry:
16710 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
16711 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
16712 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
16713 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
16714 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
16715 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
16716 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
16717 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
16718 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16719 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
16720 // CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
16721 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
16722 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16723 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
16724 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
16725 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
16726 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
16727 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
16728 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..12 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
16729 // CHECK20-NEXT:    ret void
16730 //
16731 //
16732 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..12
16733 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
16734 // CHECK20-NEXT:  entry:
16735 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16736 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16737 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
16738 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
16739 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
16740 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
16741 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
16742 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
16743 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16744 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
16745 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
16746 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16747 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16748 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
16749 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16750 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16751 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
16752 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
16753 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16754 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
16755 // CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
16756 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
16757 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16758 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
16759 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
16760 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
16761 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
16762 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16763 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16764 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16765 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
16766 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16767 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16768 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
16769 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16770 // CHECK20:       cond.true:
16771 // CHECK20-NEXT:    br label [[COND_END:%.*]]
16772 // CHECK20:       cond.false:
16773 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16774 // CHECK20-NEXT:    br label [[COND_END]]
16775 // CHECK20:       cond.end:
16776 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
16777 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16778 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16779 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
16780 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
16781 // CHECK20:       omp.inner.for.cond:
16782 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16783 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16784 // CHECK20-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
16785 // CHECK20-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16786 // CHECK20:       omp.inner.for.body:
16787 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16788 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
16789 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16790 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
16791 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
16792 // CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
16793 // CHECK20-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
16794 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
16795 // CHECK20-NEXT:    store double [[ADD4]], double* [[A]], align 4
16796 // CHECK20-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
16797 // CHECK20-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
16798 // CHECK20-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
16799 // CHECK20-NEXT:    store double [[INC]], double* [[A5]], align 4
16800 // CHECK20-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
16801 // CHECK20-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
16802 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
16803 // CHECK20-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
16804 // CHECK20-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
16805 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16806 // CHECK20:       omp.body.continue:
16807 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16808 // CHECK20:       omp.inner.for.inc:
16809 // CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16810 // CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
16811 // CHECK20-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
16812 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
16813 // CHECK20:       omp.inner.for.end:
16814 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
16815 // CHECK20:       omp.loop.exit:
16816 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
16817 // CHECK20-NEXT:    ret void
16818 //
16819 //
16820 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
16821 // CHECK20-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
16822 // CHECK20-NEXT:  entry:
16823 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16824 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16825 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16826 // CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
16827 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
16828 // CHECK20-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
16829 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16830 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16831 // CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
16832 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16833 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16834 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16835 // CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
16836 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
16837 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16838 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
16839 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
16840 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
16841 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
16842 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
16843 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
16844 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
16845 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
16846 // CHECK20-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
16847 // CHECK20-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16848 // CHECK20-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
16849 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
16850 // CHECK20-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
16851 // CHECK20-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
16852 // CHECK20-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
16853 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
16854 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
16855 // CHECK20-NEXT:    ret void
16856 //
16857 //
16858 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..15
16859 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
16860 // CHECK20-NEXT:  entry:
16861 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16862 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16863 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
16864 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16865 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16866 // CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
16867 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
16868 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
16869 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
16870 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
16871 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
16872 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
16873 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
16874 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
16875 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
16876 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16877 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16878 // CHECK20-NEXT:    [[I6:%.*]] = alloca i32, align 4
16879 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16880 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16881 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
16882 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16883 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16884 // CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
16885 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
16886 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16887 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
16888 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
16889 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
16890 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
16891 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
16892 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
16893 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16894 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16895 // CHECK20-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
16896 // CHECK20-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
16897 // CHECK20-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
16898 // CHECK20-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
16899 // CHECK20-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
16900 // CHECK20-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
16901 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16902 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
16903 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16904 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16905 // CHECK20-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
16906 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
16907 // CHECK20:       omp.precond.then:
16908 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
16909 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
16910 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
16911 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16912 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16913 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16914 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
16915 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16916 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16917 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
16918 // CHECK20-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
16919 // CHECK20-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16920 // CHECK20:       cond.true:
16921 // CHECK20-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
16922 // CHECK20-NEXT:    br label [[COND_END:%.*]]
16923 // CHECK20:       cond.false:
16924 // CHECK20-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16925 // CHECK20-NEXT:    br label [[COND_END]]
16926 // CHECK20:       cond.end:
16927 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
16928 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16929 // CHECK20-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16930 // CHECK20-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
16931 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
16932 // CHECK20:       omp.inner.for.cond:
16933 // CHECK20-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16934 // CHECK20-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16935 // CHECK20-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
16936 // CHECK20-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
16937 // CHECK20-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16938 // CHECK20:       omp.inner.for.body:
16939 // CHECK20-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16940 // CHECK20-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16941 // CHECK20-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
16942 // CHECK20-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
16943 // CHECK20-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
16944 // CHECK20-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
16945 // CHECK20-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
16946 // CHECK20-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
16947 // CHECK20-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
16948 // CHECK20-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
16949 // CHECK20-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
16950 // CHECK20-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
16951 // CHECK20-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
16952 // CHECK20-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
16953 // CHECK20-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
16954 // CHECK20-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
16955 // CHECK20-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
16956 // CHECK20-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
16957 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
16958 // CHECK20-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
16959 // CHECK20-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
16960 // CHECK20-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
16961 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
16962 // CHECK20:       omp.body.continue:
16963 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
16964 // CHECK20:       omp.inner.for.inc:
16965 // CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16966 // CHECK20-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
16967 // CHECK20-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
16968 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
16969 // CHECK20:       omp.inner.for.end:
16970 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
16971 // CHECK20:       omp.loop.exit:
16972 // CHECK20-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16973 // CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
16974 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
16975 // CHECK20-NEXT:    br label [[OMP_PRECOND_END]]
16976 // CHECK20:       omp.precond.end:
16977 // CHECK20-NEXT:    ret void
16978 //
16979 //
16980 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
16981 // CHECK20-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
16982 // CHECK20-NEXT:  entry:
16983 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
16984 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
16985 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
16986 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
16987 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
16988 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
16989 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
16990 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
16991 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
16992 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
16993 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
16994 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
16995 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
16996 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
16997 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
16998 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
16999 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
17000 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
17001 // CHECK20-NEXT:    ret void
17002 //
17003 //
17004 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..18
17005 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
17006 // CHECK20-NEXT:  entry:
17007 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
17008 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
17009 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
17010 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
17011 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
17012 // CHECK20-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17013 // CHECK20-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17014 // CHECK20-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17015 // CHECK20-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17016 // CHECK20-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17017 // CHECK20-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17018 // CHECK20-NEXT:    [[I:%.*]] = alloca i32, align 4
17019 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
17020 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
17021 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
17022 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
17023 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
17024 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
17025 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
17026 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17027 // CHECK20-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17028 // CHECK20-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17029 // CHECK20-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17030 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
17031 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
17032 // CHECK20-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17033 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17034 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
17035 // CHECK20-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17036 // CHECK20:       cond.true:
17037 // CHECK20-NEXT:    br label [[COND_END:%.*]]
17038 // CHECK20:       cond.false:
17039 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17040 // CHECK20-NEXT:    br label [[COND_END]]
17041 // CHECK20:       cond.end:
17042 // CHECK20-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
17043 // CHECK20-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17044 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17045 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
17046 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17047 // CHECK20:       omp.inner.for.cond:
17048 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17049 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17050 // CHECK20-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
17051 // CHECK20-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17052 // CHECK20:       omp.inner.for.body:
17053 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17054 // CHECK20-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
17055 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17056 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17057 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
17058 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
17059 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
17060 // CHECK20-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
17061 // CHECK20-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
17062 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
17063 // CHECK20-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
17064 // CHECK20-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
17065 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
17066 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
17067 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
17068 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
17069 // CHECK20-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17070 // CHECK20:       omp.body.continue:
17071 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17072 // CHECK20:       omp.inner.for.inc:
17073 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17074 // CHECK20-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
17075 // CHECK20-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
17076 // CHECK20-NEXT:    br label [[OMP_INNER_FOR_COND]]
17077 // CHECK20:       omp.inner.for.end:
17078 // CHECK20-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17079 // CHECK20:       omp.loop.exit:
17080 // CHECK20-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
17081 // CHECK20-NEXT:    ret void
17082 //
17083 //
17084 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
17085 // CHECK20-SAME: () #[[ATTR4]] {
17086 // CHECK20-NEXT:  entry:
17087 // CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
17088 // CHECK20-NEXT:    ret void
17089 //
17090 //
17091 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
17092 // CHECK25-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
17093 // CHECK25-NEXT:  entry:
17094 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17095 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
17096 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
17097 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17098 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
17099 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17100 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
17101 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
17102 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17103 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
17104 // CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
17105 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
17106 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
17107 // CHECK25-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
17108 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
17109 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17110 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
17111 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17112 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
17113 // CHECK25-NEXT:    ret void
17114 //
17115 //
17116 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
17117 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17118 // CHECK25-NEXT:  entry:
17119 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17120 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17121 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17122 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17123 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17124 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17125 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17126 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17127 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17128 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17129 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17130 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17131 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17132 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17133 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17134 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17135 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17136 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17137 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17138 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
17139 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17140 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17141 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
17142 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17143 // CHECK25:       cond.true:
17144 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17145 // CHECK25:       cond.false:
17146 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17147 // CHECK25-NEXT:    br label [[COND_END]]
17148 // CHECK25:       cond.end:
17149 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
17150 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17151 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17152 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
17153 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17154 // CHECK25:       omp.inner.for.cond:
17155 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17156 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17157 // CHECK25-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
17158 // CHECK25-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17159 // CHECK25:       omp.inner.for.body:
17160 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17161 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
17162 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17163 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17164 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17165 // CHECK25:       omp.body.continue:
17166 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17167 // CHECK25:       omp.inner.for.inc:
17168 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17169 // CHECK25-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
17170 // CHECK25-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
17171 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17172 // CHECK25:       omp.inner.for.end:
17173 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17174 // CHECK25:       omp.loop.exit:
17175 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
17176 // CHECK25-NEXT:    ret void
17177 //
17178 //
17179 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
17180 // CHECK25-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17181 // CHECK25-NEXT:  entry:
17182 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17183 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17184 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17185 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17186 // CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
17187 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17188 // CHECK25-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
17189 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17190 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
17191 // CHECK25-NEXT:    ret void
17192 //
17193 //
17194 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
17195 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17196 // CHECK25-NEXT:  entry:
17197 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17198 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17199 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17200 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17201 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17202 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17203 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17204 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17205 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17206 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17207 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17208 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17209 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17210 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17211 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17212 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17213 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17214 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17215 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17216 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
17217 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17218 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17219 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
17220 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17221 // CHECK25:       cond.true:
17222 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17223 // CHECK25:       cond.false:
17224 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17225 // CHECK25-NEXT:    br label [[COND_END]]
17226 // CHECK25:       cond.end:
17227 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
17228 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17229 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17230 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
17231 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17232 // CHECK25:       omp.inner.for.cond:
17233 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17234 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17235 // CHECK25-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
17236 // CHECK25-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17237 // CHECK25:       omp.inner.for.body:
17238 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17239 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
17240 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17241 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17242 // CHECK25-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
17243 // CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
17244 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
17245 // CHECK25-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
17246 // CHECK25-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
17247 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17248 // CHECK25:       omp.body.continue:
17249 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17250 // CHECK25:       omp.inner.for.inc:
17251 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17252 // CHECK25-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
17253 // CHECK25-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
17254 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17255 // CHECK25:       omp.inner.for.end:
17256 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17257 // CHECK25:       omp.loop.exit:
17258 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
17259 // CHECK25-NEXT:    ret void
17260 //
17261 //
17262 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
17263 // CHECK25-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17264 // CHECK25-NEXT:  entry:
17265 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17266 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17267 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
17268 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17269 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17270 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17271 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17272 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17273 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
17274 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
17275 // CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
17276 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
17277 // CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
17278 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17279 // CHECK25-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
17280 // CHECK25-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17281 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
17282 // CHECK25-NEXT:    ret void
17283 //
17284 //
17285 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
17286 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17287 // CHECK25-NEXT:  entry:
17288 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17289 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17290 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17291 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17292 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17293 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17294 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17295 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17296 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17297 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17298 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17299 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17300 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17301 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17302 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17303 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17304 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17305 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17306 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17307 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17308 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17309 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17310 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
17311 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17312 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17313 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
17314 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17315 // CHECK25:       cond.true:
17316 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17317 // CHECK25:       cond.false:
17318 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17319 // CHECK25-NEXT:    br label [[COND_END]]
17320 // CHECK25:       cond.end:
17321 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
17322 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17323 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17324 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
17325 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17326 // CHECK25:       omp.inner.for.cond:
17327 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17328 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17329 // CHECK25-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
17330 // CHECK25-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17331 // CHECK25:       omp.inner.for.body:
17332 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17333 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
17334 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17335 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17336 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
17337 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
17338 // CHECK25-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
17339 // CHECK25-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
17340 // CHECK25-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
17341 // CHECK25-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
17342 // CHECK25-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
17343 // CHECK25-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
17344 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17345 // CHECK25:       omp.body.continue:
17346 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17347 // CHECK25:       omp.inner.for.inc:
17348 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17349 // CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
17350 // CHECK25-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
17351 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17352 // CHECK25:       omp.inner.for.end:
17353 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17354 // CHECK25:       omp.loop.exit:
17355 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
17356 // CHECK25-NEXT:    ret void
17357 //
17358 //
17359 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
17360 // CHECK25-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
17361 // CHECK25-NEXT:  entry:
17362 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17363 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
17364 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
17365 // CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
17366 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
17367 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
17368 // CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
17369 // CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
17370 // CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
17371 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
17372 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
17373 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
17374 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17375 // CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
17376 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17377 // CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
17378 // CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
17379 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
17380 // CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
17381 // CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
17382 // CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
17383 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
17384 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17385 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
17386 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17387 // CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
17388 // CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
17389 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
17390 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
17391 // CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
17392 // CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
17393 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
17394 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
17395 // CHECK25-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
17396 // CHECK25-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
17397 // CHECK25-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
17398 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
17399 // CHECK25-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
17400 // CHECK25-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
17401 // CHECK25-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
17402 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
17403 // CHECK25-NEXT:    ret void
17404 //
17405 //
17406 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
17407 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
17408 // CHECK25-NEXT:  entry:
17409 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17410 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17411 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17412 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
17413 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
17414 // CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
17415 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
17416 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
17417 // CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
17418 // CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
17419 // CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
17420 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
17421 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17422 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17423 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17424 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17425 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17426 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17427 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17428 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17429 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17430 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17431 // CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
17432 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17433 // CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
17434 // CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
17435 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
17436 // CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
17437 // CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
17438 // CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
17439 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
17440 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17441 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
17442 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17443 // CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
17444 // CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
17445 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
17446 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
17447 // CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
17448 // CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
17449 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
17450 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17451 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17452 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17453 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17454 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
17455 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17456 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
17457 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
17458 // CHECK25-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
17459 // CHECK25:       omp.dispatch.cond:
17460 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17461 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
17462 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17463 // CHECK25:       cond.true:
17464 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17465 // CHECK25:       cond.false:
17466 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17467 // CHECK25-NEXT:    br label [[COND_END]]
17468 // CHECK25:       cond.end:
17469 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
17470 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17471 // CHECK25-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17472 // CHECK25-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
17473 // CHECK25-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17474 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17475 // CHECK25-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
17476 // CHECK25-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
17477 // CHECK25:       omp.dispatch.body:
17478 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17479 // CHECK25:       omp.inner.for.cond:
17480 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
17481 // CHECK25-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
17482 // CHECK25-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
17483 // CHECK25-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17484 // CHECK25:       omp.inner.for.body:
17485 // CHECK25-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
17486 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
17487 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17488 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
17489 // CHECK25-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !13
17490 // CHECK25-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
17491 // CHECK25-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !13
17492 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
17493 // CHECK25-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !13
17494 // CHECK25-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
17495 // CHECK25-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
17496 // CHECK25-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
17497 // CHECK25-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !13
17498 // CHECK25-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
17499 // CHECK25-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
17500 // CHECK25-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
17501 // CHECK25-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
17502 // CHECK25-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
17503 // CHECK25-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
17504 // CHECK25-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
17505 // CHECK25-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
17506 // CHECK25-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
17507 // CHECK25-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
17508 // CHECK25-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
17509 // CHECK25-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
17510 // CHECK25-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
17511 // CHECK25-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
17512 // CHECK25-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
17513 // CHECK25-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
17514 // CHECK25-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
17515 // CHECK25-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
17516 // CHECK25-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !13
17517 // CHECK25-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
17518 // CHECK25-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !13
17519 // CHECK25-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
17520 // CHECK25-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !13
17521 // CHECK25-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
17522 // CHECK25-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
17523 // CHECK25-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
17524 // CHECK25-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !13
17525 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17526 // CHECK25:       omp.body.continue:
17527 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17528 // CHECK25:       omp.inner.for.inc:
17529 // CHECK25-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
17530 // CHECK25-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
17531 // CHECK25-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
17532 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
17533 // CHECK25:       omp.inner.for.end:
17534 // CHECK25-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
17535 // CHECK25:       omp.dispatch.inc:
17536 // CHECK25-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17537 // CHECK25-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
17538 // CHECK25-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
17539 // CHECK25-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
17540 // CHECK25-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17541 // CHECK25-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
17542 // CHECK25-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
17543 // CHECK25-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
17544 // CHECK25-NEXT:    br label [[OMP_DISPATCH_COND]]
17545 // CHECK25:       omp.dispatch.end:
17546 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
17547 // CHECK25-NEXT:    ret void
17548 //
17549 //
17550 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
17551 // CHECK25-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
17552 // CHECK25-NEXT:  entry:
17553 // CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
17554 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17555 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17556 // CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
17557 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
17558 // CHECK25-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
17559 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
17560 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17561 // CHECK25-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
17562 // CHECK25-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
17563 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17564 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17565 // CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
17566 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
17567 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
17568 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17569 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17570 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
17571 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
17572 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
17573 // CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
17574 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
17575 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
17576 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
17577 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
17578 // CHECK25-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
17579 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
17580 // CHECK25-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
17581 // CHECK25-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17582 // CHECK25-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
17583 // CHECK25-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17584 // CHECK25-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
17585 // CHECK25-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
17586 // CHECK25-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
17587 // CHECK25-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
17588 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
17589 // CHECK25-NEXT:    ret void
17590 //
17591 //
17592 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
17593 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
17594 // CHECK25-NEXT:  entry:
17595 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17596 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17597 // CHECK25-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
17598 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17599 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17600 // CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
17601 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
17602 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17603 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17604 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17605 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
17606 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
17607 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17608 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17609 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17610 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17611 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17612 // CHECK25-NEXT:    [[I8:%.*]] = alloca i32, align 4
17613 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17614 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17615 // CHECK25-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
17616 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17617 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17618 // CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
17619 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
17620 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
17621 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17622 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17623 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
17624 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
17625 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
17626 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
17627 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
17628 // CHECK25-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
17629 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
17630 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17631 // CHECK25-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
17632 // CHECK25-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
17633 // CHECK25-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
17634 // CHECK25-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
17635 // CHECK25-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
17636 // CHECK25-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
17637 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17638 // CHECK25-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
17639 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17640 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
17641 // CHECK25-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
17642 // CHECK25-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17643 // CHECK25:       omp.precond.then:
17644 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17645 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
17646 // CHECK25-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
17647 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17648 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17649 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17650 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
17651 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17652 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17653 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
17654 // CHECK25-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
17655 // CHECK25-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17656 // CHECK25:       cond.true:
17657 // CHECK25-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
17658 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17659 // CHECK25:       cond.false:
17660 // CHECK25-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17661 // CHECK25-NEXT:    br label [[COND_END]]
17662 // CHECK25:       cond.end:
17663 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
17664 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17665 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17666 // CHECK25-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
17667 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17668 // CHECK25:       omp.inner.for.cond:
17669 // CHECK25-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17670 // CHECK25-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17671 // CHECK25-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
17672 // CHECK25-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
17673 // CHECK25-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17674 // CHECK25:       omp.inner.for.body:
17675 // CHECK25-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17676 // CHECK25-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17677 // CHECK25-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
17678 // CHECK25-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
17679 // CHECK25-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
17680 // CHECK25-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
17681 // CHECK25-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
17682 // CHECK25-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
17683 // CHECK25-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
17684 // CHECK25-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
17685 // CHECK25-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
17686 // CHECK25-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
17687 // CHECK25-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
17688 // CHECK25-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
17689 // CHECK25-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
17690 // CHECK25-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
17691 // CHECK25-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
17692 // CHECK25-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
17693 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
17694 // CHECK25-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
17695 // CHECK25-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
17696 // CHECK25-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
17697 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17698 // CHECK25:       omp.body.continue:
17699 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17700 // CHECK25:       omp.inner.for.inc:
17701 // CHECK25-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17702 // CHECK25-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
17703 // CHECK25-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
17704 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17705 // CHECK25:       omp.inner.for.end:
17706 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17707 // CHECK25:       omp.loop.exit:
17708 // CHECK25-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17709 // CHECK25-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
17710 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
17711 // CHECK25-NEXT:    br label [[OMP_PRECOND_END]]
17712 // CHECK25:       omp.precond.end:
17713 // CHECK25-NEXT:    ret void
17714 //
17715 //
17716 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
17717 // CHECK25-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
17718 // CHECK25-NEXT:  entry:
17719 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
17720 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
17721 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
17722 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
17723 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
17724 // CHECK25-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
17725 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
17726 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
17727 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17728 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
17729 // CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
17730 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
17731 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
17732 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17733 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
17734 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
17735 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
17736 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
17737 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
17738 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
17739 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
17740 // CHECK25-NEXT:    ret void
17741 //
17742 //
17743 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
17744 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
17745 // CHECK25-NEXT:  entry:
17746 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17747 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17748 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
17749 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
17750 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
17751 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
17752 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
17753 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17754 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17755 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17756 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17757 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17758 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17759 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17760 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17761 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17762 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
17763 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
17764 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17765 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
17766 // CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
17767 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
17768 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
17769 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17770 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
17771 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
17772 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17773 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17774 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17775 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17776 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17777 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
17778 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17779 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17780 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
17781 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17782 // CHECK25:       cond.true:
17783 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17784 // CHECK25:       cond.false:
17785 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17786 // CHECK25-NEXT:    br label [[COND_END]]
17787 // CHECK25:       cond.end:
17788 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
17789 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17790 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17791 // CHECK25-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
17792 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17793 // CHECK25:       omp.inner.for.cond:
17794 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17795 // CHECK25-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17796 // CHECK25-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
17797 // CHECK25-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17798 // CHECK25:       omp.inner.for.body:
17799 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17800 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
17801 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17802 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17803 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
17804 // CHECK25-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
17805 // CHECK25-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
17806 // CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
17807 // CHECK25-NEXT:    store double [[ADD5]], double* [[A]], align 8
17808 // CHECK25-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
17809 // CHECK25-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
17810 // CHECK25-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
17811 // CHECK25-NEXT:    store double [[INC]], double* [[A6]], align 8
17812 // CHECK25-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
17813 // CHECK25-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
17814 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
17815 // CHECK25-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
17816 // CHECK25-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
17817 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17818 // CHECK25:       omp.body.continue:
17819 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17820 // CHECK25:       omp.inner.for.inc:
17821 // CHECK25-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17822 // CHECK25-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
17823 // CHECK25-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
17824 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17825 // CHECK25:       omp.inner.for.end:
17826 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17827 // CHECK25:       omp.loop.exit:
17828 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
17829 // CHECK25-NEXT:    ret void
17830 //
17831 //
17832 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
17833 // CHECK25-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
17834 // CHECK25-NEXT:  entry:
17835 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17836 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17837 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
17838 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
17839 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17840 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17841 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17842 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
17843 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17844 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17845 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
17846 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
17847 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
17848 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
17849 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
17850 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
17851 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17852 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
17853 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17854 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
17855 // CHECK25-NEXT:    ret void
17856 //
17857 //
17858 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..6
17859 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
17860 // CHECK25-NEXT:  entry:
17861 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17862 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17863 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
17864 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17865 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
17866 // CHECK25-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17867 // CHECK25-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17868 // CHECK25-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17869 // CHECK25-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17870 // CHECK25-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17871 // CHECK25-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17872 // CHECK25-NEXT:    [[I:%.*]] = alloca i32, align 4
17873 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17874 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17875 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
17876 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17877 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
17878 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
17879 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17880 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
17881 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17882 // CHECK25-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17883 // CHECK25-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17884 // CHECK25-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17885 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17886 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
17887 // CHECK25-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17888 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17889 // CHECK25-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
17890 // CHECK25-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17891 // CHECK25:       cond.true:
17892 // CHECK25-NEXT:    br label [[COND_END:%.*]]
17893 // CHECK25:       cond.false:
17894 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17895 // CHECK25-NEXT:    br label [[COND_END]]
17896 // CHECK25:       cond.end:
17897 // CHECK25-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
17898 // CHECK25-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17899 // CHECK25-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17900 // CHECK25-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
17901 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
17902 // CHECK25:       omp.inner.for.cond:
17903 // CHECK25-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17904 // CHECK25-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17905 // CHECK25-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
17906 // CHECK25-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17907 // CHECK25:       omp.inner.for.body:
17908 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17909 // CHECK25-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
17910 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17911 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
17912 // CHECK25-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
17913 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
17914 // CHECK25-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
17915 // CHECK25-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
17916 // CHECK25-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
17917 // CHECK25-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
17918 // CHECK25-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
17919 // CHECK25-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
17920 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
17921 // CHECK25-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
17922 // CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
17923 // CHECK25-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
17924 // CHECK25-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
17925 // CHECK25:       omp.body.continue:
17926 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
17927 // CHECK25:       omp.inner.for.inc:
17928 // CHECK25-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17929 // CHECK25-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
17930 // CHECK25-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
17931 // CHECK25-NEXT:    br label [[OMP_INNER_FOR_COND]]
17932 // CHECK25:       omp.inner.for.end:
17933 // CHECK25-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
17934 // CHECK25:       omp.loop.exit:
17935 // CHECK25-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
17936 // CHECK25-NEXT:    ret void
17937 //
17938 //
17939 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
17940 // CHECK26-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
17941 // CHECK26-NEXT:  entry:
17942 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17943 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
17944 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
17945 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
17946 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
17947 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17948 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
17949 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
17950 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17951 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
17952 // CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
17953 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
17954 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
17955 // CHECK26-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
17956 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
17957 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
17958 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
17959 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
17960 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
17961 // CHECK26-NEXT:    ret void
17962 //
17963 //
17964 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
17965 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
17966 // CHECK26-NEXT:  entry:
17967 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17968 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17969 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
17970 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
17971 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
17972 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
17973 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
17974 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17975 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17976 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
17977 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17978 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17979 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
17980 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
17981 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
17982 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
17983 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17984 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17985 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17986 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
17987 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17988 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17989 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
17990 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17991 // CHECK26:       cond.true:
17992 // CHECK26-NEXT:    br label [[COND_END:%.*]]
17993 // CHECK26:       cond.false:
17994 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17995 // CHECK26-NEXT:    br label [[COND_END]]
17996 // CHECK26:       cond.end:
17997 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
17998 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17999 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18000 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
18001 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18002 // CHECK26:       omp.inner.for.cond:
18003 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18004 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18005 // CHECK26-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
18006 // CHECK26-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18007 // CHECK26:       omp.inner.for.body:
18008 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18009 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
18010 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18011 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18012 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18013 // CHECK26:       omp.body.continue:
18014 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18015 // CHECK26:       omp.inner.for.inc:
18016 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18017 // CHECK26-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
18018 // CHECK26-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
18019 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18020 // CHECK26:       omp.inner.for.end:
18021 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18022 // CHECK26:       omp.loop.exit:
18023 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
18024 // CHECK26-NEXT:    ret void
18025 //
18026 //
18027 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
18028 // CHECK26-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
18029 // CHECK26-NEXT:  entry:
18030 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18031 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
18032 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18033 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18034 // CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
18035 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
18036 // CHECK26-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
18037 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
18038 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
18039 // CHECK26-NEXT:    ret void
18040 //
18041 //
18042 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
18043 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
18044 // CHECK26-NEXT:  entry:
18045 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18046 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18047 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18048 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18049 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18050 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18051 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18052 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18053 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18054 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18055 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18056 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18057 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18058 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18059 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18060 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18061 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18062 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18063 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18064 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
18065 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18066 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18067 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
18068 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18069 // CHECK26:       cond.true:
18070 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18071 // CHECK26:       cond.false:
18072 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18073 // CHECK26-NEXT:    br label [[COND_END]]
18074 // CHECK26:       cond.end:
18075 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
18076 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18077 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18078 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
18079 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18080 // CHECK26:       omp.inner.for.cond:
18081 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18082 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18083 // CHECK26-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
18084 // CHECK26-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18085 // CHECK26:       omp.inner.for.body:
18086 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18087 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
18088 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18089 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18090 // CHECK26-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
18091 // CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
18092 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
18093 // CHECK26-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
18094 // CHECK26-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
18095 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18096 // CHECK26:       omp.body.continue:
18097 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18098 // CHECK26:       omp.inner.for.inc:
18099 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18100 // CHECK26-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
18101 // CHECK26-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
18102 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18103 // CHECK26:       omp.inner.for.end:
18104 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18105 // CHECK26:       omp.loop.exit:
18106 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
18107 // CHECK26-NEXT:    ret void
18108 //
18109 //
18110 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
18111 // CHECK26-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
18112 // CHECK26-NEXT:  entry:
18113 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18114 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18115 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
18116 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
18117 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18118 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18119 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18120 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18121 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
18122 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
18123 // CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
18124 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
18125 // CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
18126 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
18127 // CHECK26-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
18128 // CHECK26-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
18129 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
18130 // CHECK26-NEXT:    ret void
18131 //
18132 //
18133 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
18134 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
18135 // CHECK26-NEXT:  entry:
18136 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18137 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18138 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18139 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18140 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18141 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18142 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18143 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18144 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18145 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18146 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18147 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18148 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18149 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18150 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18151 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18152 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18153 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18154 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18155 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18156 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18157 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18158 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
18159 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18160 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18161 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
18162 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18163 // CHECK26:       cond.true:
18164 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18165 // CHECK26:       cond.false:
18166 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18167 // CHECK26-NEXT:    br label [[COND_END]]
18168 // CHECK26:       cond.end:
18169 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
18170 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18171 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18172 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
18173 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18174 // CHECK26:       omp.inner.for.cond:
18175 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18176 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18177 // CHECK26-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
18178 // CHECK26-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18179 // CHECK26:       omp.inner.for.body:
18180 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18181 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
18182 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18183 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18184 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
18185 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
18186 // CHECK26-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
18187 // CHECK26-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2
18188 // CHECK26-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
18189 // CHECK26-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
18190 // CHECK26-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
18191 // CHECK26-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
18192 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18193 // CHECK26:       omp.body.continue:
18194 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18195 // CHECK26:       omp.inner.for.inc:
18196 // CHECK26-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18197 // CHECK26-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
18198 // CHECK26-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
18199 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18200 // CHECK26:       omp.inner.for.end:
18201 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18202 // CHECK26:       omp.loop.exit:
18203 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
18204 // CHECK26-NEXT:    ret void
18205 //
18206 //
18207 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
18208 // CHECK26-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
18209 // CHECK26-NEXT:  entry:
18210 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18211 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
18212 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
18213 // CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
18214 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
18215 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
18216 // CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
18217 // CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
18218 // CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
18219 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
18220 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
18221 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
18222 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18223 // CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
18224 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18225 // CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
18226 // CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
18227 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
18228 // CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
18229 // CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
18230 // CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
18231 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
18232 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18233 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
18234 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18235 // CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
18236 // CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
18237 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
18238 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
18239 // CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
18240 // CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
18241 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
18242 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
18243 // CHECK26-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
18244 // CHECK26-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
18245 // CHECK26-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
18246 // CHECK26-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
18247 // CHECK26-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
18248 // CHECK26-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
18249 // CHECK26-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
18250 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
18251 // CHECK26-NEXT:    ret void
18252 //
18253 //
18254 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
18255 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
18256 // CHECK26-NEXT:  entry:
18257 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18258 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18259 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18260 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
18261 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
18262 // CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
18263 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
18264 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
18265 // CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
18266 // CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
18267 // CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
18268 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
18269 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18270 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18271 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18272 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18273 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18274 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18275 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18276 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18277 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18278 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18279 // CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
18280 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18281 // CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
18282 // CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
18283 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
18284 // CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
18285 // CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
18286 // CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
18287 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
18288 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18289 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
18290 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18291 // CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
18292 // CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
18293 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
18294 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
18295 // CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
18296 // CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
18297 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
18298 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18299 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18300 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18301 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18302 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
18303 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18304 // CHECK26-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
18305 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
18306 // CHECK26-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
18307 // CHECK26:       omp.dispatch.cond:
18308 // CHECK26-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18309 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
18310 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18311 // CHECK26:       cond.true:
18312 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18313 // CHECK26:       cond.false:
18314 // CHECK26-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18315 // CHECK26-NEXT:    br label [[COND_END]]
18316 // CHECK26:       cond.end:
18317 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
18318 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18319 // CHECK26-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18320 // CHECK26-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
18321 // CHECK26-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18322 // CHECK26-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18323 // CHECK26-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
18324 // CHECK26-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
18325 // CHECK26:       omp.dispatch.body:
18326 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18327 // CHECK26:       omp.inner.for.cond:
18328 // CHECK26-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
18329 // CHECK26-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
18330 // CHECK26-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
18331 // CHECK26-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18332 // CHECK26:       omp.inner.for.body:
18333 // CHECK26-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
18334 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
18335 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18336 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
18337 // CHECK26-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !13
18338 // CHECK26-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP19]], 1
18339 // CHECK26-NEXT:    store i32 [[ADD8]], i32* [[CONV]], align 4, !llvm.access.group !13
18340 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
18341 // CHECK26-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !13
18342 // CHECK26-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
18343 // CHECK26-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
18344 // CHECK26-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
18345 // CHECK26-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !13
18346 // CHECK26-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
18347 // CHECK26-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
18348 // CHECK26-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
18349 // CHECK26-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
18350 // CHECK26-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
18351 // CHECK26-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !13
18352 // CHECK26-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
18353 // CHECK26-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
18354 // CHECK26-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
18355 // CHECK26-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
18356 // CHECK26-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !13
18357 // CHECK26-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
18358 // CHECK26-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
18359 // CHECK26-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
18360 // CHECK26-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
18361 // CHECK26-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
18362 // CHECK26-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !13
18363 // CHECK26-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
18364 // CHECK26-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !13
18365 // CHECK26-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
18366 // CHECK26-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !13
18367 // CHECK26-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
18368 // CHECK26-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !13
18369 // CHECK26-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
18370 // CHECK26-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
18371 // CHECK26-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
18372 // CHECK26-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !13
18373 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18374 // CHECK26:       omp.body.continue:
18375 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18376 // CHECK26:       omp.inner.for.inc:
18377 // CHECK26-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
18378 // CHECK26-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
18379 // CHECK26-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
18380 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
18381 // CHECK26:       omp.inner.for.end:
18382 // CHECK26-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
18383 // CHECK26:       omp.dispatch.inc:
18384 // CHECK26-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18385 // CHECK26-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18386 // CHECK26-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
18387 // CHECK26-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
18388 // CHECK26-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18389 // CHECK26-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18390 // CHECK26-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
18391 // CHECK26-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
18392 // CHECK26-NEXT:    br label [[OMP_DISPATCH_COND]]
18393 // CHECK26:       omp.dispatch.end:
18394 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
18395 // CHECK26-NEXT:    ret void
18396 //
18397 //
18398 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
18399 // CHECK26-SAME: (i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
18400 // CHECK26-NEXT:  entry:
18401 // CHECK26-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
18402 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18403 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18404 // CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
18405 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
18406 // CHECK26-NEXT:    [[N_CASTED:%.*]] = alloca i64, align 8
18407 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
18408 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
18409 // CHECK26-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
18410 // CHECK26-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
18411 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18412 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18413 // CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
18414 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
18415 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
18416 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18417 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18418 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
18419 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
18420 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
18421 // CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
18422 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV4]], align 4
18423 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[N_CASTED]], align 8
18424 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
18425 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
18426 // CHECK26-NEXT:    store i32 [[TMP3]], i32* [[CONV5]], align 4
18427 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[A_CASTED]], align 8
18428 // CHECK26-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV2]], align 2
18429 // CHECK26-NEXT:    [[CONV6:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
18430 // CHECK26-NEXT:    store i16 [[TMP5]], i16* [[CONV6]], align 2
18431 // CHECK26-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AA_CASTED]], align 8
18432 // CHECK26-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
18433 // CHECK26-NEXT:    [[CONV7:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
18434 // CHECK26-NEXT:    store i8 [[TMP7]], i8* [[CONV7]], align 1
18435 // CHECK26-NEXT:    [[TMP8:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
18436 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP8]], [10 x i32]* [[TMP0]])
18437 // CHECK26-NEXT:    ret void
18438 //
18439 //
18440 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
18441 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
18442 // CHECK26-NEXT:  entry:
18443 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18444 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18445 // CHECK26-NEXT:    [[N_ADDR:%.*]] = alloca i64, align 8
18446 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18447 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18448 // CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
18449 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
18450 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18451 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18452 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
18453 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
18454 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
18455 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18456 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18457 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18458 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18459 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18460 // CHECK26-NEXT:    [[I8:%.*]] = alloca i32, align 4
18461 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18462 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18463 // CHECK26-NEXT:    store i64 [[N]], i64* [[N_ADDR]], align 8
18464 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18465 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18466 // CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
18467 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
18468 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
18469 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18470 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18471 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
18472 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
18473 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV1]], align 4
18474 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
18475 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
18476 // CHECK26-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_4]], align 4
18477 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
18478 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18479 // CHECK26-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
18480 // CHECK26-NEXT:    [[SUB6:%.*]] = sub i32 [[SUB]], 1
18481 // CHECK26-NEXT:    [[ADD:%.*]] = add i32 [[SUB6]], 1
18482 // CHECK26-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
18483 // CHECK26-NEXT:    [[SUB7:%.*]] = sub i32 [[DIV]], 1
18484 // CHECK26-NEXT:    store i32 [[SUB7]], i32* [[DOTCAPTURE_EXPR_5]], align 4
18485 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18486 // CHECK26-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
18487 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18488 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
18489 // CHECK26-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
18490 // CHECK26-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18491 // CHECK26:       omp.precond.then:
18492 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18493 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
18494 // CHECK26-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
18495 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18496 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18497 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18498 // CHECK26-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
18499 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18500 // CHECK26-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18501 // CHECK26-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
18502 // CHECK26-NEXT:    [[CMP9:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
18503 // CHECK26-NEXT:    br i1 [[CMP9]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18504 // CHECK26:       cond.true:
18505 // CHECK26-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
18506 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18507 // CHECK26:       cond.false:
18508 // CHECK26-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18509 // CHECK26-NEXT:    br label [[COND_END]]
18510 // CHECK26:       cond.end:
18511 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
18512 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18513 // CHECK26-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18514 // CHECK26-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
18515 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18516 // CHECK26:       omp.inner.for.cond:
18517 // CHECK26-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18518 // CHECK26-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18519 // CHECK26-NEXT:    [[ADD10:%.*]] = add i32 [[TMP17]], 1
18520 // CHECK26-NEXT:    [[CMP11:%.*]] = icmp ult i32 [[TMP16]], [[ADD10]]
18521 // CHECK26-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18522 // CHECK26:       omp.inner.for.body:
18523 // CHECK26-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18524 // CHECK26-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18525 // CHECK26-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
18526 // CHECK26-NEXT:    [[ADD12:%.*]] = add i32 [[TMP18]], [[MUL]]
18527 // CHECK26-NEXT:    store i32 [[ADD12]], i32* [[I8]], align 4
18528 // CHECK26-NEXT:    [[TMP20:%.*]] = load i32, i32* [[CONV1]], align 4
18529 // CHECK26-NEXT:    [[ADD13:%.*]] = add nsw i32 [[TMP20]], 1
18530 // CHECK26-NEXT:    store i32 [[ADD13]], i32* [[CONV1]], align 4
18531 // CHECK26-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV2]], align 2
18532 // CHECK26-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP21]] to i32
18533 // CHECK26-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
18534 // CHECK26-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
18535 // CHECK26-NEXT:    store i16 [[CONV16]], i16* [[CONV2]], align 2
18536 // CHECK26-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV3]], align 1
18537 // CHECK26-NEXT:    [[CONV17:%.*]] = sext i8 [[TMP22]] to i32
18538 // CHECK26-NEXT:    [[ADD18:%.*]] = add nsw i32 [[CONV17]], 1
18539 // CHECK26-NEXT:    [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
18540 // CHECK26-NEXT:    store i8 [[CONV19]], i8* [[CONV3]], align 1
18541 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
18542 // CHECK26-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
18543 // CHECK26-NEXT:    [[ADD20:%.*]] = add nsw i32 [[TMP23]], 1
18544 // CHECK26-NEXT:    store i32 [[ADD20]], i32* [[ARRAYIDX]], align 4
18545 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18546 // CHECK26:       omp.body.continue:
18547 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18548 // CHECK26:       omp.inner.for.inc:
18549 // CHECK26-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18550 // CHECK26-NEXT:    [[ADD21:%.*]] = add i32 [[TMP24]], 1
18551 // CHECK26-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
18552 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18553 // CHECK26:       omp.inner.for.end:
18554 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18555 // CHECK26:       omp.loop.exit:
18556 // CHECK26-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18557 // CHECK26-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
18558 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
18559 // CHECK26-NEXT:    br label [[OMP_PRECOND_END]]
18560 // CHECK26:       omp.precond.end:
18561 // CHECK26-NEXT:    ret void
18562 //
18563 //
18564 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
18565 // CHECK26-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
18566 // CHECK26-NEXT:  entry:
18567 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
18568 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
18569 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
18570 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
18571 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
18572 // CHECK26-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
18573 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
18574 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
18575 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18576 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
18577 // CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
18578 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
18579 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
18580 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18581 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
18582 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
18583 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
18584 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
18585 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
18586 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
18587 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
18588 // CHECK26-NEXT:    ret void
18589 //
18590 //
18591 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
18592 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
18593 // CHECK26-NEXT:  entry:
18594 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18595 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18596 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
18597 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
18598 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
18599 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
18600 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
18601 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18602 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18603 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18604 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18605 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18606 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18607 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18608 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18609 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18610 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
18611 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
18612 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18613 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
18614 // CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
18615 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
18616 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
18617 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18618 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
18619 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
18620 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18621 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18622 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18623 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18624 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18625 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
18626 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18627 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18628 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
18629 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18630 // CHECK26:       cond.true:
18631 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18632 // CHECK26:       cond.false:
18633 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18634 // CHECK26-NEXT:    br label [[COND_END]]
18635 // CHECK26:       cond.end:
18636 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
18637 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18638 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18639 // CHECK26-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
18640 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18641 // CHECK26:       omp.inner.for.cond:
18642 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18643 // CHECK26-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18644 // CHECK26-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
18645 // CHECK26-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18646 // CHECK26:       omp.inner.for.body:
18647 // CHECK26-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18648 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
18649 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18650 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18651 // CHECK26-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
18652 // CHECK26-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
18653 // CHECK26-NEXT:    [[ADD5:%.*]] = fadd double [[CONV4]], 1.500000e+00
18654 // CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
18655 // CHECK26-NEXT:    store double [[ADD5]], double* [[A]], align 8
18656 // CHECK26-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
18657 // CHECK26-NEXT:    [[TMP13:%.*]] = load double, double* [[A6]], align 8
18658 // CHECK26-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
18659 // CHECK26-NEXT:    store double [[INC]], double* [[A6]], align 8
18660 // CHECK26-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
18661 // CHECK26-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
18662 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
18663 // CHECK26-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
18664 // CHECK26-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2
18665 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18666 // CHECK26:       omp.body.continue:
18667 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18668 // CHECK26:       omp.inner.for.inc:
18669 // CHECK26-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18670 // CHECK26-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP15]], 1
18671 // CHECK26-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
18672 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18673 // CHECK26:       omp.inner.for.end:
18674 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18675 // CHECK26:       omp.loop.exit:
18676 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
18677 // CHECK26-NEXT:    ret void
18678 //
18679 //
18680 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
18681 // CHECK26-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
18682 // CHECK26-NEXT:  entry:
18683 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18684 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18685 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
18686 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
18687 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
18688 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18689 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18690 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
18691 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18692 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18693 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
18694 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
18695 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
18696 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
18697 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
18698 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
18699 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
18700 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
18701 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
18702 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
18703 // CHECK26-NEXT:    ret void
18704 //
18705 //
18706 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..6
18707 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
18708 // CHECK26-NEXT:  entry:
18709 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18710 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18711 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
18712 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
18713 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
18714 // CHECK26-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18715 // CHECK26-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18716 // CHECK26-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18717 // CHECK26-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18718 // CHECK26-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18719 // CHECK26-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18720 // CHECK26-NEXT:    [[I:%.*]] = alloca i32, align 4
18721 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18722 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18723 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
18724 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
18725 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
18726 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
18727 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
18728 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
18729 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18730 // CHECK26-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18731 // CHECK26-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18732 // CHECK26-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18733 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18734 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
18735 // CHECK26-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18736 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18737 // CHECK26-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
18738 // CHECK26-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18739 // CHECK26:       cond.true:
18740 // CHECK26-NEXT:    br label [[COND_END:%.*]]
18741 // CHECK26:       cond.false:
18742 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18743 // CHECK26-NEXT:    br label [[COND_END]]
18744 // CHECK26:       cond.end:
18745 // CHECK26-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
18746 // CHECK26-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18747 // CHECK26-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18748 // CHECK26-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
18749 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18750 // CHECK26:       omp.inner.for.cond:
18751 // CHECK26-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18752 // CHECK26-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18753 // CHECK26-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
18754 // CHECK26-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18755 // CHECK26:       omp.inner.for.body:
18756 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18757 // CHECK26-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
18758 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18759 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18760 // CHECK26-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4
18761 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
18762 // CHECK26-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4
18763 // CHECK26-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2
18764 // CHECK26-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
18765 // CHECK26-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
18766 // CHECK26-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
18767 // CHECK26-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2
18768 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
18769 // CHECK26-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
18770 // CHECK26-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
18771 // CHECK26-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4
18772 // CHECK26-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18773 // CHECK26:       omp.body.continue:
18774 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18775 // CHECK26:       omp.inner.for.inc:
18776 // CHECK26-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18777 // CHECK26-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP12]], 1
18778 // CHECK26-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
18779 // CHECK26-NEXT:    br label [[OMP_INNER_FOR_COND]]
18780 // CHECK26:       omp.inner.for.end:
18781 // CHECK26-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18782 // CHECK26:       omp.loop.exit:
18783 // CHECK26-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
18784 // CHECK26-NEXT:    ret void
18785 //
18786 //
18787 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
18788 // CHECK27-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
18789 // CHECK27-NEXT:  entry:
18790 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18791 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
18792 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
18793 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
18794 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
18795 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18796 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
18797 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
18798 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18799 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
18800 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
18801 // CHECK27-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
18802 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
18803 // CHECK27-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
18804 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
18805 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
18806 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
18807 // CHECK27-NEXT:    ret void
18808 //
18809 //
18810 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
18811 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
18812 // CHECK27-NEXT:  entry:
18813 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
18814 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
18815 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18816 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18817 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18818 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18819 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18820 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18821 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18822 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
18823 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
18824 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
18825 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18826 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18827 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18828 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18829 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18830 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18831 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
18832 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
18833 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18834 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18835 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
18836 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18837 // CHECK27:       cond.true:
18838 // CHECK27-NEXT:    br label [[COND_END:%.*]]
18839 // CHECK27:       cond.false:
18840 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18841 // CHECK27-NEXT:    br label [[COND_END]]
18842 // CHECK27:       cond.end:
18843 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
18844 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18845 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18846 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
18847 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18848 // CHECK27:       omp.inner.for.cond:
18849 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18850 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18851 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
18852 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18853 // CHECK27:       omp.inner.for.body:
18854 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18855 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
18856 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18857 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18858 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18859 // CHECK27:       omp.body.continue:
18860 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18861 // CHECK27:       omp.inner.for.inc:
18862 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18863 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
18864 // CHECK27-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
18865 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
18866 // CHECK27:       omp.inner.for.end:
18867 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18868 // CHECK27:       omp.loop.exit:
18869 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
18870 // CHECK27-NEXT:    ret void
18871 //
18872 //
18873 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
18874 // CHECK27-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
18875 // CHECK27-NEXT:  entry:
18876 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18877 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
18878 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18879 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18880 // CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
18881 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
18882 // CHECK27-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
18883 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
18884 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
18885 // CHECK27-NEXT:    ret void
18886 //
18887 //
18888 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
18889 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
18890 // CHECK27-NEXT:  entry:
18891 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
18892 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
18893 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18894 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18895 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18896 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18897 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18898 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18899 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18900 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
18901 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
18902 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
18903 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18904 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18905 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18906 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18907 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18908 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18909 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
18910 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
18911 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18912 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18913 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
18914 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18915 // CHECK27:       cond.true:
18916 // CHECK27-NEXT:    br label [[COND_END:%.*]]
18917 // CHECK27:       cond.false:
18918 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18919 // CHECK27-NEXT:    br label [[COND_END]]
18920 // CHECK27:       cond.end:
18921 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
18922 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18923 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18924 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
18925 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
18926 // CHECK27:       omp.inner.for.cond:
18927 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18928 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18929 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
18930 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18931 // CHECK27:       omp.inner.for.body:
18932 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18933 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
18934 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18935 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
18936 // CHECK27-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
18937 // CHECK27-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
18938 // CHECK27-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
18939 // CHECK27-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
18940 // CHECK27-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
18941 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
18942 // CHECK27:       omp.body.continue:
18943 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
18944 // CHECK27:       omp.inner.for.inc:
18945 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18946 // CHECK27-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
18947 // CHECK27-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
18948 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
18949 // CHECK27:       omp.inner.for.end:
18950 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
18951 // CHECK27:       omp.loop.exit:
18952 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
18953 // CHECK27-NEXT:    ret void
18954 //
18955 //
18956 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
18957 // CHECK27-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
18958 // CHECK27-NEXT:  entry:
18959 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
18960 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18961 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
18962 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
18963 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
18964 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18965 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18966 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
18967 // CHECK27-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
18968 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
18969 // CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
18970 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
18971 // CHECK27-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
18972 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
18973 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
18974 // CHECK27-NEXT:    ret void
18975 //
18976 //
18977 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
18978 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
18979 // CHECK27-NEXT:  entry:
18980 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
18981 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
18982 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
18983 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
18984 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
18985 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
18986 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
18987 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
18988 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18989 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18990 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
18991 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
18992 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
18993 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
18994 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
18995 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
18996 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
18997 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
18998 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18999 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19000 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19001 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
19002 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19003 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19004 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
19005 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19006 // CHECK27:       cond.true:
19007 // CHECK27-NEXT:    br label [[COND_END:%.*]]
19008 // CHECK27:       cond.false:
19009 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19010 // CHECK27-NEXT:    br label [[COND_END]]
19011 // CHECK27:       cond.end:
19012 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
19013 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19014 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19015 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
19016 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19017 // CHECK27:       omp.inner.for.cond:
19018 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19019 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19020 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
19021 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19022 // CHECK27:       omp.inner.for.body:
19023 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19024 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
19025 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19026 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19027 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
19028 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
19029 // CHECK27-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
19030 // CHECK27-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
19031 // CHECK27-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
19032 // CHECK27-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
19033 // CHECK27-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
19034 // CHECK27-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
19035 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19036 // CHECK27:       omp.body.continue:
19037 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19038 // CHECK27:       omp.inner.for.inc:
19039 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19040 // CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
19041 // CHECK27-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
19042 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
19043 // CHECK27:       omp.inner.for.end:
19044 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19045 // CHECK27:       omp.loop.exit:
19046 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
19047 // CHECK27-NEXT:    ret void
19048 //
19049 //
19050 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
19051 // CHECK27-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
19052 // CHECK27-NEXT:  entry:
19053 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19054 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
19055 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19056 // CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
19057 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
19058 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19059 // CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
19060 // CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
19061 // CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
19062 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
19063 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
19064 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
19065 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19066 // CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
19067 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19068 // CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
19069 // CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
19070 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19071 // CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
19072 // CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
19073 // CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
19074 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19075 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
19076 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19077 // CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
19078 // CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
19079 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19080 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
19081 // CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
19082 // CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
19083 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
19084 // CHECK27-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
19085 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
19086 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19087 // CHECK27-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
19088 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
19089 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
19090 // CHECK27-NEXT:    ret void
19091 //
19092 //
19093 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
19094 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
19095 // CHECK27-NEXT:  entry:
19096 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19097 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19098 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19099 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
19100 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19101 // CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
19102 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
19103 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19104 // CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
19105 // CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
19106 // CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
19107 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
19108 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19109 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19110 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19111 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19112 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19113 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19114 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
19115 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19116 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19117 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19118 // CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
19119 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19120 // CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
19121 // CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
19122 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19123 // CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
19124 // CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
19125 // CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
19126 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19127 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
19128 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19129 // CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
19130 // CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
19131 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19132 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
19133 // CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
19134 // CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
19135 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19136 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19137 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19138 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19139 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19140 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19141 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
19142 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
19143 // CHECK27-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
19144 // CHECK27:       omp.dispatch.cond:
19145 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19146 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
19147 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19148 // CHECK27:       cond.true:
19149 // CHECK27-NEXT:    br label [[COND_END:%.*]]
19150 // CHECK27:       cond.false:
19151 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19152 // CHECK27-NEXT:    br label [[COND_END]]
19153 // CHECK27:       cond.end:
19154 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
19155 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19156 // CHECK27-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19157 // CHECK27-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
19158 // CHECK27-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19159 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19160 // CHECK27-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
19161 // CHECK27-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
19162 // CHECK27:       omp.dispatch.body:
19163 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19164 // CHECK27:       omp.inner.for.cond:
19165 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19166 // CHECK27-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
19167 // CHECK27-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
19168 // CHECK27-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19169 // CHECK27:       omp.inner.for.body:
19170 // CHECK27-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19171 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
19172 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19173 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
19174 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !14
19175 // CHECK27-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
19176 // CHECK27-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !14
19177 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
19178 // CHECK27-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
19179 // CHECK27-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
19180 // CHECK27-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
19181 // CHECK27-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
19182 // CHECK27-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !14
19183 // CHECK27-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
19184 // CHECK27-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
19185 // CHECK27-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
19186 // CHECK27-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
19187 // CHECK27-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
19188 // CHECK27-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
19189 // CHECK27-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
19190 // CHECK27-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
19191 // CHECK27-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
19192 // CHECK27-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
19193 // CHECK27-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
19194 // CHECK27-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
19195 // CHECK27-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
19196 // CHECK27-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
19197 // CHECK27-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
19198 // CHECK27-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
19199 // CHECK27-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
19200 // CHECK27-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
19201 // CHECK27-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !14
19202 // CHECK27-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
19203 // CHECK27-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !14
19204 // CHECK27-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
19205 // CHECK27-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !14
19206 // CHECK27-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
19207 // CHECK27-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
19208 // CHECK27-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
19209 // CHECK27-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !14
19210 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19211 // CHECK27:       omp.body.continue:
19212 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19213 // CHECK27:       omp.inner.for.inc:
19214 // CHECK27-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19215 // CHECK27-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
19216 // CHECK27-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19217 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
19218 // CHECK27:       omp.inner.for.end:
19219 // CHECK27-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
19220 // CHECK27:       omp.dispatch.inc:
19221 // CHECK27-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19222 // CHECK27-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19223 // CHECK27-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
19224 // CHECK27-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
19225 // CHECK27-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19226 // CHECK27-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19227 // CHECK27-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
19228 // CHECK27-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
19229 // CHECK27-NEXT:    br label [[OMP_DISPATCH_COND]]
19230 // CHECK27:       omp.dispatch.end:
19231 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
19232 // CHECK27-NEXT:    ret void
19233 //
19234 //
19235 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
19236 // CHECK27-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
19237 // CHECK27-NEXT:  entry:
19238 // CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
19239 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19240 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19241 // CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
19242 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
19243 // CHECK27-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
19244 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
19245 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
19246 // CHECK27-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
19247 // CHECK27-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
19248 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19249 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19250 // CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
19251 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
19252 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19253 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
19254 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
19255 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
19256 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
19257 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
19258 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
19259 // CHECK27-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
19260 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
19261 // CHECK27-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
19262 // CHECK27-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
19263 // CHECK27-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
19264 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
19265 // CHECK27-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
19266 // CHECK27-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
19267 // CHECK27-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
19268 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
19269 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
19270 // CHECK27-NEXT:    ret void
19271 //
19272 //
19273 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
19274 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
19275 // CHECK27-NEXT:  entry:
19276 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19277 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19278 // CHECK27-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
19279 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19280 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19281 // CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
19282 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
19283 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19284 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19285 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
19286 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
19287 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
19288 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
19289 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19290 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19291 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19292 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19293 // CHECK27-NEXT:    [[I6:%.*]] = alloca i32, align 4
19294 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19295 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19296 // CHECK27-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
19297 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19298 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19299 // CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
19300 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
19301 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19302 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
19303 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
19304 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
19305 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
19306 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
19307 // CHECK27-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
19308 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
19309 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19310 // CHECK27-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
19311 // CHECK27-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
19312 // CHECK27-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
19313 // CHECK27-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
19314 // CHECK27-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
19315 // CHECK27-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
19316 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19317 // CHECK27-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
19318 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19319 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
19320 // CHECK27-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
19321 // CHECK27-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
19322 // CHECK27:       omp.precond.then:
19323 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19324 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
19325 // CHECK27-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
19326 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19327 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19328 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19329 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
19330 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19331 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19332 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
19333 // CHECK27-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
19334 // CHECK27-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19335 // CHECK27:       cond.true:
19336 // CHECK27-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
19337 // CHECK27-NEXT:    br label [[COND_END:%.*]]
19338 // CHECK27:       cond.false:
19339 // CHECK27-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19340 // CHECK27-NEXT:    br label [[COND_END]]
19341 // CHECK27:       cond.end:
19342 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
19343 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19344 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19345 // CHECK27-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
19346 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19347 // CHECK27:       omp.inner.for.cond:
19348 // CHECK27-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19349 // CHECK27-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19350 // CHECK27-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
19351 // CHECK27-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
19352 // CHECK27-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19353 // CHECK27:       omp.inner.for.body:
19354 // CHECK27-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19355 // CHECK27-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19356 // CHECK27-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
19357 // CHECK27-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
19358 // CHECK27-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
19359 // CHECK27-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
19360 // CHECK27-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
19361 // CHECK27-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
19362 // CHECK27-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
19363 // CHECK27-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
19364 // CHECK27-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
19365 // CHECK27-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
19366 // CHECK27-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
19367 // CHECK27-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
19368 // CHECK27-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
19369 // CHECK27-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
19370 // CHECK27-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
19371 // CHECK27-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
19372 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
19373 // CHECK27-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
19374 // CHECK27-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
19375 // CHECK27-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
19376 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19377 // CHECK27:       omp.body.continue:
19378 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19379 // CHECK27:       omp.inner.for.inc:
19380 // CHECK27-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19381 // CHECK27-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
19382 // CHECK27-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
19383 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
19384 // CHECK27:       omp.inner.for.end:
19385 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19386 // CHECK27:       omp.loop.exit:
19387 // CHECK27-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19388 // CHECK27-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
19389 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
19390 // CHECK27-NEXT:    br label [[OMP_PRECOND_END]]
19391 // CHECK27:       omp.precond.end:
19392 // CHECK27-NEXT:    ret void
19393 //
19394 //
19395 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
19396 // CHECK27-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
19397 // CHECK27-NEXT:  entry:
19398 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
19399 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
19400 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19401 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19402 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
19403 // CHECK27-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
19404 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
19405 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
19406 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19407 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19408 // CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
19409 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
19410 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19411 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19412 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
19413 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
19414 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
19415 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
19416 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
19417 // CHECK27-NEXT:    ret void
19418 //
19419 //
19420 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
19421 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
19422 // CHECK27-NEXT:  entry:
19423 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19424 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19425 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
19426 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
19427 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19428 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19429 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
19430 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19431 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19432 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19433 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19434 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19435 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19436 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
19437 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19438 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19439 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
19440 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
19441 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19442 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19443 // CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
19444 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
19445 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19446 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19447 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
19448 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19449 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19450 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19451 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19452 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19453 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
19454 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19455 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19456 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
19457 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19458 // CHECK27:       cond.true:
19459 // CHECK27-NEXT:    br label [[COND_END:%.*]]
19460 // CHECK27:       cond.false:
19461 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19462 // CHECK27-NEXT:    br label [[COND_END]]
19463 // CHECK27:       cond.end:
19464 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
19465 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19466 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19467 // CHECK27-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
19468 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19469 // CHECK27:       omp.inner.for.cond:
19470 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19471 // CHECK27-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19472 // CHECK27-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
19473 // CHECK27-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19474 // CHECK27:       omp.inner.for.body:
19475 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19476 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
19477 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19478 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19479 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
19480 // CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
19481 // CHECK27-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
19482 // CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
19483 // CHECK27-NEXT:    store double [[ADD4]], double* [[A]], align 4
19484 // CHECK27-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
19485 // CHECK27-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
19486 // CHECK27-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
19487 // CHECK27-NEXT:    store double [[INC]], double* [[A5]], align 4
19488 // CHECK27-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
19489 // CHECK27-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
19490 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
19491 // CHECK27-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
19492 // CHECK27-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
19493 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19494 // CHECK27:       omp.body.continue:
19495 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19496 // CHECK27:       omp.inner.for.inc:
19497 // CHECK27-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19498 // CHECK27-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
19499 // CHECK27-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
19500 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
19501 // CHECK27:       omp.inner.for.end:
19502 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19503 // CHECK27:       omp.loop.exit:
19504 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
19505 // CHECK27-NEXT:    ret void
19506 //
19507 //
19508 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
19509 // CHECK27-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
19510 // CHECK27-NEXT:  entry:
19511 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19512 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19513 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
19514 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
19515 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
19516 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19517 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19518 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
19519 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19520 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
19521 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
19522 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
19523 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
19524 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
19525 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
19526 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
19527 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
19528 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
19529 // CHECK27-NEXT:    ret void
19530 //
19531 //
19532 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..6
19533 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
19534 // CHECK27-NEXT:  entry:
19535 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19536 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19537 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19538 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19539 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
19540 // CHECK27-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19541 // CHECK27-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19542 // CHECK27-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19543 // CHECK27-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19544 // CHECK27-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19545 // CHECK27-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19546 // CHECK27-NEXT:    [[I:%.*]] = alloca i32, align 4
19547 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19548 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19549 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19550 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19551 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
19552 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19553 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
19554 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19555 // CHECK27-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19556 // CHECK27-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19557 // CHECK27-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19558 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19559 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
19560 // CHECK27-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19561 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19562 // CHECK27-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
19563 // CHECK27-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19564 // CHECK27:       cond.true:
19565 // CHECK27-NEXT:    br label [[COND_END:%.*]]
19566 // CHECK27:       cond.false:
19567 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19568 // CHECK27-NEXT:    br label [[COND_END]]
19569 // CHECK27:       cond.end:
19570 // CHECK27-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
19571 // CHECK27-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19572 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19573 // CHECK27-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
19574 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19575 // CHECK27:       omp.inner.for.cond:
19576 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19577 // CHECK27-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19578 // CHECK27-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
19579 // CHECK27-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19580 // CHECK27:       omp.inner.for.body:
19581 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19582 // CHECK27-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
19583 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19584 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19585 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
19586 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
19587 // CHECK27-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
19588 // CHECK27-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
19589 // CHECK27-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
19590 // CHECK27-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
19591 // CHECK27-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
19592 // CHECK27-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
19593 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
19594 // CHECK27-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
19595 // CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
19596 // CHECK27-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
19597 // CHECK27-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19598 // CHECK27:       omp.body.continue:
19599 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19600 // CHECK27:       omp.inner.for.inc:
19601 // CHECK27-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19602 // CHECK27-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
19603 // CHECK27-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
19604 // CHECK27-NEXT:    br label [[OMP_INNER_FOR_COND]]
19605 // CHECK27:       omp.inner.for.end:
19606 // CHECK27-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19607 // CHECK27:       omp.loop.exit:
19608 // CHECK27-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
19609 // CHECK27-NEXT:    ret void
19610 //
19611 //
19612 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l103
19613 // CHECK28-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
19614 // CHECK28-NEXT:  entry:
19615 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19616 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
19617 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
19618 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
19619 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
19620 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19621 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19622 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
19623 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19624 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19625 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
19626 // CHECK28-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
19627 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
19628 // CHECK28-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
19629 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
19630 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
19631 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
19632 // CHECK28-NEXT:    ret void
19633 //
19634 //
19635 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
19636 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
19637 // CHECK28-NEXT:  entry:
19638 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19639 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19640 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19641 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19642 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19643 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19644 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19645 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19646 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19647 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
19648 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19649 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19650 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19651 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19652 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19653 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19654 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19655 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19656 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19657 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
19658 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19659 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19660 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
19661 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19662 // CHECK28:       cond.true:
19663 // CHECK28-NEXT:    br label [[COND_END:%.*]]
19664 // CHECK28:       cond.false:
19665 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19666 // CHECK28-NEXT:    br label [[COND_END]]
19667 // CHECK28:       cond.end:
19668 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
19669 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19670 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19671 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
19672 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19673 // CHECK28:       omp.inner.for.cond:
19674 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19675 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19676 // CHECK28-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
19677 // CHECK28-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19678 // CHECK28:       omp.inner.for.body:
19679 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19680 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
19681 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19682 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19683 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19684 // CHECK28:       omp.body.continue:
19685 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19686 // CHECK28:       omp.inner.for.inc:
19687 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19688 // CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
19689 // CHECK28-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
19690 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
19691 // CHECK28:       omp.inner.for.end:
19692 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19693 // CHECK28:       omp.loop.exit:
19694 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
19695 // CHECK28-NEXT:    ret void
19696 //
19697 //
19698 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l113
19699 // CHECK28-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
19700 // CHECK28-NEXT:  entry:
19701 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19702 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
19703 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19704 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19705 // CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
19706 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
19707 // CHECK28-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
19708 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
19709 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
19710 // CHECK28-NEXT:    ret void
19711 //
19712 //
19713 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
19714 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
19715 // CHECK28-NEXT:  entry:
19716 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19717 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19718 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19719 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19720 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19721 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19722 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19723 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19724 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19725 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
19726 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19727 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19728 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19729 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19730 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19731 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19732 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19733 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19734 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19735 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
19736 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19737 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19738 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
19739 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19740 // CHECK28:       cond.true:
19741 // CHECK28-NEXT:    br label [[COND_END:%.*]]
19742 // CHECK28:       cond.false:
19743 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19744 // CHECK28-NEXT:    br label [[COND_END]]
19745 // CHECK28:       cond.end:
19746 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
19747 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19748 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19749 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
19750 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19751 // CHECK28:       omp.inner.for.cond:
19752 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19753 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19754 // CHECK28-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
19755 // CHECK28-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19756 // CHECK28:       omp.inner.for.body:
19757 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19758 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
19759 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19760 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19761 // CHECK28-NEXT:    [[TMP8:%.*]] = load i16, i16* [[CONV]], align 2
19762 // CHECK28-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP8]] to i32
19763 // CHECK28-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
19764 // CHECK28-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
19765 // CHECK28-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
19766 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19767 // CHECK28:       omp.body.continue:
19768 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19769 // CHECK28:       omp.inner.for.inc:
19770 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19771 // CHECK28-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP9]], 1
19772 // CHECK28-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4
19773 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
19774 // CHECK28:       omp.inner.for.end:
19775 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19776 // CHECK28:       omp.loop.exit:
19777 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
19778 // CHECK28-NEXT:    ret void
19779 //
19780 //
19781 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l120
19782 // CHECK28-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
19783 // CHECK28-NEXT:  entry:
19784 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19785 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19786 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
19787 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
19788 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19789 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19790 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19791 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
19792 // CHECK28-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
19793 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
19794 // CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
19795 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
19796 // CHECK28-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
19797 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
19798 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
19799 // CHECK28-NEXT:    ret void
19800 //
19801 //
19802 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
19803 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
19804 // CHECK28-NEXT:  entry:
19805 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19806 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19807 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19808 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
19809 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19810 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19811 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19812 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19813 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19814 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19815 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
19816 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19817 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19818 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19819 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
19820 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
19821 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19822 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19823 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19824 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19825 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19826 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
19827 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19828 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19829 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 9
19830 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19831 // CHECK28:       cond.true:
19832 // CHECK28-NEXT:    br label [[COND_END:%.*]]
19833 // CHECK28:       cond.false:
19834 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19835 // CHECK28-NEXT:    br label [[COND_END]]
19836 // CHECK28:       cond.end:
19837 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
19838 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19839 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19840 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
19841 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19842 // CHECK28:       omp.inner.for.cond:
19843 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19844 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19845 // CHECK28-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
19846 // CHECK28-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19847 // CHECK28:       omp.inner.for.body:
19848 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19849 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
19850 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19851 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
19852 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
19853 // CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
19854 // CHECK28-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
19855 // CHECK28-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2
19856 // CHECK28-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP9]] to i32
19857 // CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
19858 // CHECK28-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
19859 // CHECK28-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
19860 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
19861 // CHECK28:       omp.body.continue:
19862 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
19863 // CHECK28:       omp.inner.for.inc:
19864 // CHECK28-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19865 // CHECK28-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP10]], 1
19866 // CHECK28-NEXT:    store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
19867 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
19868 // CHECK28:       omp.inner.for.end:
19869 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
19870 // CHECK28:       omp.loop.exit:
19871 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
19872 // CHECK28-NEXT:    ret void
19873 //
19874 //
19875 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l145
19876 // CHECK28-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
19877 // CHECK28-NEXT:  entry:
19878 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19879 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
19880 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19881 // CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
19882 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
19883 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19884 // CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
19885 // CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
19886 // CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
19887 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
19888 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
19889 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
19890 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19891 // CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
19892 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19893 // CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
19894 // CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
19895 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19896 // CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
19897 // CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
19898 // CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
19899 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19900 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
19901 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19902 // CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
19903 // CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
19904 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19905 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
19906 // CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
19907 // CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
19908 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
19909 // CHECK28-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
19910 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
19911 // CHECK28-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19912 // CHECK28-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
19913 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
19914 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
19915 // CHECK28-NEXT:    ret void
19916 //
19917 //
19918 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
19919 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
19920 // CHECK28-NEXT:  entry:
19921 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
19922 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
19923 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
19924 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
19925 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
19926 // CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
19927 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
19928 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
19929 // CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
19930 // CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
19931 // CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
19932 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
19933 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
19934 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
19935 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
19936 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
19937 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19938 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19939 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
19940 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
19941 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
19942 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
19943 // CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
19944 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
19945 // CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
19946 // CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
19947 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
19948 // CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
19949 // CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
19950 // CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
19951 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19952 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
19953 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
19954 // CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
19955 // CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
19956 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
19957 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
19958 // CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
19959 // CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
19960 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
19961 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
19962 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19963 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19964 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
19965 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
19966 // CHECK28-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
19967 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
19968 // CHECK28-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
19969 // CHECK28:       omp.dispatch.cond:
19970 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19971 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 9
19972 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19973 // CHECK28:       cond.true:
19974 // CHECK28-NEXT:    br label [[COND_END:%.*]]
19975 // CHECK28:       cond.false:
19976 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19977 // CHECK28-NEXT:    br label [[COND_END]]
19978 // CHECK28:       cond.end:
19979 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
19980 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19981 // CHECK28-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19982 // CHECK28-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
19983 // CHECK28-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19984 // CHECK28-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19985 // CHECK28-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
19986 // CHECK28-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
19987 // CHECK28:       omp.dispatch.body:
19988 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
19989 // CHECK28:       omp.inner.for.cond:
19990 // CHECK28-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19991 // CHECK28-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
19992 // CHECK28-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
19993 // CHECK28-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19994 // CHECK28:       omp.inner.for.body:
19995 // CHECK28-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
19996 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
19997 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19998 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
19999 // CHECK28-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !14
20000 // CHECK28-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
20001 // CHECK28-NEXT:    store i32 [[ADD7]], i32* [[A_ADDR]], align 4, !llvm.access.group !14
20002 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
20003 // CHECK28-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
20004 // CHECK28-NEXT:    [[CONV:%.*]] = fpext float [[TMP20]] to double
20005 // CHECK28-NEXT:    [[ADD8:%.*]] = fadd double [[CONV]], 1.000000e+00
20006 // CHECK28-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
20007 // CHECK28-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !14
20008 // CHECK28-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
20009 // CHECK28-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
20010 // CHECK28-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
20011 // CHECK28-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
20012 // CHECK28-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
20013 // CHECK28-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
20014 // CHECK28-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
20015 // CHECK28-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
20016 // CHECK28-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
20017 // CHECK28-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
20018 // CHECK28-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !14
20019 // CHECK28-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
20020 // CHECK28-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
20021 // CHECK28-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
20022 // CHECK28-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
20023 // CHECK28-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
20024 // CHECK28-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !14
20025 // CHECK28-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
20026 // CHECK28-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !14
20027 // CHECK28-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
20028 // CHECK28-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !14
20029 // CHECK28-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
20030 // CHECK28-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !14
20031 // CHECK28-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
20032 // CHECK28-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
20033 // CHECK28-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
20034 // CHECK28-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !14
20035 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
20036 // CHECK28:       omp.body.continue:
20037 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
20038 // CHECK28:       omp.inner.for.inc:
20039 // CHECK28-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
20040 // CHECK28-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
20041 // CHECK28-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
20042 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
20043 // CHECK28:       omp.inner.for.end:
20044 // CHECK28-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
20045 // CHECK28:       omp.dispatch.inc:
20046 // CHECK28-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20047 // CHECK28-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20048 // CHECK28-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
20049 // CHECK28-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
20050 // CHECK28-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20051 // CHECK28-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20052 // CHECK28-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
20053 // CHECK28-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
20054 // CHECK28-NEXT:    br label [[OMP_DISPATCH_COND]]
20055 // CHECK28:       omp.dispatch.end:
20056 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
20057 // CHECK28-NEXT:    ret void
20058 //
20059 //
20060 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l200
20061 // CHECK28-SAME: (i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
20062 // CHECK28-NEXT:  entry:
20063 // CHECK28-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
20064 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
20065 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
20066 // CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
20067 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
20068 // CHECK28-NEXT:    [[N_CASTED:%.*]] = alloca i32, align 4
20069 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
20070 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
20071 // CHECK28-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
20072 // CHECK28-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
20073 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
20074 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
20075 // CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
20076 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
20077 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
20078 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
20079 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
20080 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
20081 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[N_CASTED]], align 4
20082 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_CASTED]], align 4
20083 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A_ADDR]], align 4
20084 // CHECK28-NEXT:    store i32 [[TMP3]], i32* [[A_CASTED]], align 4
20085 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_CASTED]], align 4
20086 // CHECK28-NEXT:    [[TMP5:%.*]] = load i16, i16* [[CONV]], align 2
20087 // CHECK28-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
20088 // CHECK28-NEXT:    store i16 [[TMP5]], i16* [[CONV2]], align 2
20089 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AA_CASTED]], align 4
20090 // CHECK28-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV1]], align 1
20091 // CHECK28-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
20092 // CHECK28-NEXT:    store i8 [[TMP7]], i8* [[CONV3]], align 1
20093 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
20094 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP8]], [10 x i32]* [[TMP0]])
20095 // CHECK28-NEXT:    ret void
20096 //
20097 //
20098 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
20099 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
20100 // CHECK28-NEXT:  entry:
20101 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
20102 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
20103 // CHECK28-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
20104 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
20105 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
20106 // CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
20107 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
20108 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
20109 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
20110 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20111 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
20112 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
20113 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
20114 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
20115 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
20116 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20117 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20118 // CHECK28-NEXT:    [[I6:%.*]] = alloca i32, align 4
20119 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
20120 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
20121 // CHECK28-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
20122 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
20123 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
20124 // CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
20125 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
20126 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
20127 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
20128 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
20129 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
20130 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
20131 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
20132 // CHECK28-NEXT:    store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_2]], align 4
20133 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20134 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20135 // CHECK28-NEXT:    [[SUB:%.*]] = sub i32 [[TMP3]], [[TMP4]]
20136 // CHECK28-NEXT:    [[SUB4:%.*]] = sub i32 [[SUB]], 1
20137 // CHECK28-NEXT:    [[ADD:%.*]] = add i32 [[SUB4]], 1
20138 // CHECK28-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
20139 // CHECK28-NEXT:    [[SUB5:%.*]] = sub i32 [[DIV]], 1
20140 // CHECK28-NEXT:    store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
20141 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20142 // CHECK28-NEXT:    store i32 [[TMP5]], i32* [[I]], align 4
20143 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20144 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20145 // CHECK28-NEXT:    [[CMP:%.*]] = icmp slt i32 [[TMP6]], [[TMP7]]
20146 // CHECK28-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20147 // CHECK28:       omp.precond.then:
20148 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
20149 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
20150 // CHECK28-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
20151 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20152 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20153 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
20154 // CHECK28-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
20155 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20156 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20157 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
20158 // CHECK28-NEXT:    [[CMP7:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
20159 // CHECK28-NEXT:    br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20160 // CHECK28:       cond.true:
20161 // CHECK28-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
20162 // CHECK28-NEXT:    br label [[COND_END:%.*]]
20163 // CHECK28:       cond.false:
20164 // CHECK28-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20165 // CHECK28-NEXT:    br label [[COND_END]]
20166 // CHECK28:       cond.end:
20167 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
20168 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20169 // CHECK28-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20170 // CHECK28-NEXT:    store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
20171 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
20172 // CHECK28:       omp.inner.for.cond:
20173 // CHECK28-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20174 // CHECK28-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20175 // CHECK28-NEXT:    [[ADD8:%.*]] = add i32 [[TMP17]], 1
20176 // CHECK28-NEXT:    [[CMP9:%.*]] = icmp ult i32 [[TMP16]], [[ADD8]]
20177 // CHECK28-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20178 // CHECK28:       omp.inner.for.body:
20179 // CHECK28-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20180 // CHECK28-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20181 // CHECK28-NEXT:    [[MUL:%.*]] = mul i32 [[TMP19]], 1
20182 // CHECK28-NEXT:    [[ADD10:%.*]] = add i32 [[TMP18]], [[MUL]]
20183 // CHECK28-NEXT:    store i32 [[ADD10]], i32* [[I6]], align 4
20184 // CHECK28-NEXT:    [[TMP20:%.*]] = load i32, i32* [[A_ADDR]], align 4
20185 // CHECK28-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP20]], 1
20186 // CHECK28-NEXT:    store i32 [[ADD11]], i32* [[A_ADDR]], align 4
20187 // CHECK28-NEXT:    [[TMP21:%.*]] = load i16, i16* [[CONV]], align 2
20188 // CHECK28-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP21]] to i32
20189 // CHECK28-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
20190 // CHECK28-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
20191 // CHECK28-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2
20192 // CHECK28-NEXT:    [[TMP22:%.*]] = load i8, i8* [[CONV1]], align 1
20193 // CHECK28-NEXT:    [[CONV15:%.*]] = sext i8 [[TMP22]] to i32
20194 // CHECK28-NEXT:    [[ADD16:%.*]] = add nsw i32 [[CONV15]], 1
20195 // CHECK28-NEXT:    [[CONV17:%.*]] = trunc i32 [[ADD16]] to i8
20196 // CHECK28-NEXT:    store i8 [[CONV17]], i8* [[CONV1]], align 1
20197 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
20198 // CHECK28-NEXT:    [[TMP23:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
20199 // CHECK28-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP23]], 1
20200 // CHECK28-NEXT:    store i32 [[ADD18]], i32* [[ARRAYIDX]], align 4
20201 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
20202 // CHECK28:       omp.body.continue:
20203 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
20204 // CHECK28:       omp.inner.for.inc:
20205 // CHECK28-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20206 // CHECK28-NEXT:    [[ADD19:%.*]] = add i32 [[TMP24]], 1
20207 // CHECK28-NEXT:    store i32 [[ADD19]], i32* [[DOTOMP_IV]], align 4
20208 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
20209 // CHECK28:       omp.inner.for.end:
20210 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
20211 // CHECK28:       omp.loop.exit:
20212 // CHECK28-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
20213 // CHECK28-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
20214 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
20215 // CHECK28-NEXT:    br label [[OMP_PRECOND_END]]
20216 // CHECK28:       omp.precond.end:
20217 // CHECK28-NEXT:    ret void
20218 //
20219 //
20220 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l218
20221 // CHECK28-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
20222 // CHECK28-NEXT:  entry:
20223 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
20224 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
20225 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
20226 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
20227 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
20228 // CHECK28-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
20229 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
20230 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
20231 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
20232 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
20233 // CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
20234 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
20235 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
20236 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
20237 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
20238 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
20239 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
20240 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
20241 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
20242 // CHECK28-NEXT:    ret void
20243 //
20244 //
20245 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
20246 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
20247 // CHECK28-NEXT:  entry:
20248 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
20249 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
20250 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
20251 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
20252 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
20253 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
20254 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
20255 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
20256 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
20257 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
20258 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
20259 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20260 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20261 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
20262 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
20263 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
20264 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
20265 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
20266 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
20267 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
20268 // CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
20269 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
20270 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
20271 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
20272 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
20273 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
20274 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
20275 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20276 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20277 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
20278 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
20279 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20280 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20281 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
20282 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20283 // CHECK28:       cond.true:
20284 // CHECK28-NEXT:    br label [[COND_END:%.*]]
20285 // CHECK28:       cond.false:
20286 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20287 // CHECK28-NEXT:    br label [[COND_END]]
20288 // CHECK28:       cond.end:
20289 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
20290 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20291 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20292 // CHECK28-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
20293 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
20294 // CHECK28:       omp.inner.for.cond:
20295 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20296 // CHECK28-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20297 // CHECK28-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
20298 // CHECK28-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20299 // CHECK28:       omp.inner.for.body:
20300 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20301 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
20302 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20303 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
20304 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4
20305 // CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
20306 // CHECK28-NEXT:    [[ADD4:%.*]] = fadd double [[CONV]], 1.500000e+00
20307 // CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
20308 // CHECK28-NEXT:    store double [[ADD4]], double* [[A]], align 4
20309 // CHECK28-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
20310 // CHECK28-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 4
20311 // CHECK28-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
20312 // CHECK28-NEXT:    store double [[INC]], double* [[A5]], align 4
20313 // CHECK28-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
20314 // CHECK28-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
20315 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
20316 // CHECK28-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
20317 // CHECK28-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2
20318 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
20319 // CHECK28:       omp.body.continue:
20320 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
20321 // CHECK28:       omp.inner.for.inc:
20322 // CHECK28-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20323 // CHECK28-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP15]], 1
20324 // CHECK28-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
20325 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
20326 // CHECK28:       omp.inner.for.end:
20327 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
20328 // CHECK28:       omp.loop.exit:
20329 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
20330 // CHECK28-NEXT:    ret void
20331 //
20332 //
20333 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l183
20334 // CHECK28-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
20335 // CHECK28-NEXT:  entry:
20336 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
20337 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
20338 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
20339 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
20340 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
20341 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
20342 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
20343 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
20344 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
20345 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
20346 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
20347 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
20348 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
20349 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
20350 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
20351 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
20352 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
20353 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
20354 // CHECK28-NEXT:    ret void
20355 //
20356 //
20357 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..6
20358 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
20359 // CHECK28-NEXT:  entry:
20360 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
20361 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
20362 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
20363 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
20364 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
20365 // CHECK28-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
20366 // CHECK28-NEXT:    [[TMP:%.*]] = alloca i32, align 4
20367 // CHECK28-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
20368 // CHECK28-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
20369 // CHECK28-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20370 // CHECK28-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20371 // CHECK28-NEXT:    [[I:%.*]] = alloca i32, align 4
20372 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
20373 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
20374 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
20375 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
20376 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
20377 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
20378 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
20379 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
20380 // CHECK28-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
20381 // CHECK28-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20382 // CHECK28-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20383 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
20384 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
20385 // CHECK28-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20386 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20387 // CHECK28-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
20388 // CHECK28-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20389 // CHECK28:       cond.true:
20390 // CHECK28-NEXT:    br label [[COND_END:%.*]]
20391 // CHECK28:       cond.false:
20392 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20393 // CHECK28-NEXT:    br label [[COND_END]]
20394 // CHECK28:       cond.end:
20395 // CHECK28-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
20396 // CHECK28-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20397 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20398 // CHECK28-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
20399 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
20400 // CHECK28:       omp.inner.for.cond:
20401 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20402 // CHECK28-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20403 // CHECK28-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
20404 // CHECK28-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20405 // CHECK28:       omp.inner.for.body:
20406 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20407 // CHECK28-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
20408 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20409 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
20410 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4
20411 // CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
20412 // CHECK28-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4
20413 // CHECK28-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2
20414 // CHECK28-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
20415 // CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
20416 // CHECK28-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
20417 // CHECK28-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2
20418 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
20419 // CHECK28-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
20420 // CHECK28-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
20421 // CHECK28-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4
20422 // CHECK28-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
20423 // CHECK28:       omp.body.continue:
20424 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
20425 // CHECK28:       omp.inner.for.inc:
20426 // CHECK28-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20427 // CHECK28-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP12]], 1
20428 // CHECK28-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
20429 // CHECK28-NEXT:    br label [[OMP_INNER_FOR_COND]]
20430 // CHECK28:       omp.inner.for.end:
20431 // CHECK28-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
20432 // CHECK28:       omp.loop.exit:
20433 // CHECK28-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
20434 // CHECK28-NEXT:    ret void
20435 //
20436