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 // 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
37 // 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
38 // 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
39 // 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
40 // 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
41 // 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
42 
43 // 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}}"
44 // 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
45 // 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}}"
46 // 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}}"
47 // 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
48 // 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}}"
49 
50 // Test target codegen - host bc file has to be created first.
51 // 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
52 // 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
53 // 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
54 // 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
55 // 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
56 // 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
57 // 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
58 // 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
59 
60 // 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
61 // 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}}"
62 // 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
63 // 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}}"
64 // 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
65 // 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}}"
66 // 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
67 // 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}}"
68 
69 // expected-no-diagnostics
70 #ifndef HEADER
71 #define HEADER
72 
73 
74 
75 
76 // We have 8 target regions, but only 6 that actually will generate offloading
77 // code and have mapped arguments, and only 4 have all-constant map sizes.
78 
79 
80 
81 // Check target registration is registered as a Ctor.
82 
83 
84 template<typename tx, typename ty>
85 struct TT{
86   tx X;
87   ty Y;
88 };
89 
90 int global;
91 
92 int foo(int n) {
93   int a = 0;
94   short aa = 0;
95   float b[10];
96   float bn[n];
97   double c[5][10];
98   double cn[5][n];
99   TT<long long, char> d;
100 
101   #pragma omp target teams num_teams(a) thread_limit(a) firstprivate(aa) nowait
102   {
103   }
104 
105   #pragma omp target teams if(target: 0)
106   {
107     a += 1;
108   }
109 
110 
111   #pragma omp target teams if(target: 1)
112   {
113     aa += 1;
114   }
115 
116 
117 
118   #pragma omp target teams if(target: n>10)
119   {
120     a += 1;
121     aa += 1;
122   }
123 
124   // We capture 3 VLA sizes in this target region
125 
126 
127 
128 
129 
130   // The names below are not necessarily consistent with the names used for the
131   // addresses above as some are repeated.
132 
133 
134 
135 
136 
137 
138 
139 
140 
141 
142   #pragma omp target teams if(target: n>20)
143   {
144     a += 1;
145     b[2] += 1.0;
146     bn[3] += 1.0;
147     c[1][2] += 1.0;
148     cn[1][3] += 1.0;
149     d.X += 1;
150     d.Y += 1;
151   }
152 
153   const int nn = 0;
154   #pragma omp target teams shared(nn)
155   #pragma omp parallel firstprivate(nn)
156   (void)nn;
157   #pragma omp target teams firstprivate(nn)
158   #pragma omp parallel shared(nn)
159   (void)nn;
160   return a;
161 }
162 
163 // Check that the offloading functions are emitted and that the arguments are
164 // correct and loaded correctly for the target regions in foo().
165 
166 
167 
168 // Create stack storage and store argument in there.
169 
170 // Create stack storage and store argument in there.
171 
172 // Create stack storage and store argument in there.
173 
174 // Create local storage for each capture.
175 
176 
177 
178 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
179 
180 
181 void bazzzz(int n, int f[n]) {
182 #pragma omp target teams private(f)
183   ;
184 }
185 
186 template<typename tx>
187 tx ftemplate(int n) {
188   tx a = 0;
189   short aa = 0;
190   tx b[10];
191 
192   #pragma omp target teams if(target: n>40)
193   {
194     a += 1;
195     aa += 1;
196     b[2] += 1;
197   }
198 
199   return a;
200 }
201 
202 static
203 int fstatic(int n) {
204   int a = 0;
205   short aa = 0;
206   char aaa = 0;
207   int b[10];
208 
209   #pragma omp target teams if(target: n>50)
210   {
211     a += 1;
212     aa += 1;
213     aaa += 1;
214     b[2] += 1;
215   }
216 
217   return a;
218 }
219 
220 struct S1 {
221   double a;
222 
223   int r1(int n){
224     int b = n+1;
225     short int c[2][n];
226 
227     #pragma omp target teams if(target: n>60)
228     {
229       this->a = (double)b + 1.5;
230       c[1][1] = ++a;
231     }
232 
233     return c[1][1] + (int)b;
234   }
235 };
236 
237 int bar(int n){
238   int a = 0;
239 
240   a += foo(n);
241 
242   S1 S;
243   a += S.r1(n);
244 
245   a += fstatic(n);
246 
247   a += ftemplate<int>(n);
248 
249   return a;
250 }
251 
252 
253 
254 // We capture 2 VLA sizes in this target region
255 
256 
257 // The names below are not necessarily consistent with the names used for the
258 // addresses above as some are repeated.
259 
260 
261 
262 
263 
264 
265 
266 
267 
268 
269 
270 
271 
272 
273 
274 
275 
276 
277 
278 
279 // Check that the offloading functions are emitted and that the arguments are
280 // correct and loaded correctly for the target regions of the callees of bar().
281 
282 // Create local storage for each capture.
283 // Store captures in the context.
284 
285 
286 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
287 
288 
289 // Create local storage for each capture.
290 // Store captures in the context.
291 
292 
293 
294 
295 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
296 
297 // Create local storage for each capture.
298 // Store captures in the context.
299 
300 
301 
302 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
303 
304 #endif
305 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooi
306 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
307 // CHECK1-NEXT:  entry:
308 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
309 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
310 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
311 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
312 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
313 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
314 // CHECK1-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
315 // CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
316 // CHECK1-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
317 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
318 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
320 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
321 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
322 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
323 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
324 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
325 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
326 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
327 // CHECK1-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
328 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
329 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
330 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
331 // CHECK1-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
332 // CHECK1-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
333 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
334 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
335 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
336 // CHECK1-NEXT:    [[A_CASTED21:%.*]] = alloca i64, align 8
337 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS25:%.*]] = alloca [9 x i8*], align 8
338 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS26:%.*]] = alloca [9 x i8*], align 8
339 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS27:%.*]] = alloca [9 x i8*], align 8
340 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
341 // CHECK1-NEXT:    [[NN:%.*]] = alloca i32, align 4
342 // CHECK1-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
343 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 8
344 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 8
345 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 8
346 // CHECK1-NEXT:    [[NN_CASTED38:%.*]] = alloca i64, align 8
347 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS40:%.*]] = alloca [1 x i8*], align 8
348 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS41:%.*]] = alloca [1 x i8*], align 8
349 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS42:%.*]] = alloca [1 x i8*], align 8
350 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
351 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
352 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
353 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
354 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
355 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
356 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
357 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
358 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
359 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
360 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
361 // CHECK1-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
362 // CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
363 // CHECK1-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
364 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
365 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
366 // CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
367 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
368 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
369 // CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
370 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
371 // CHECK1-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
372 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
373 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
374 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
375 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
376 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
377 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
378 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
379 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
380 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
381 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
382 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
383 // CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
384 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
385 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
386 // CHECK1-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
387 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
388 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
389 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
390 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
391 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
392 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
393 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
394 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
395 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
396 // CHECK1-NEXT:    store i8* null, i8** [[TMP24]], align 8
397 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
398 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
399 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
400 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
401 // CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
402 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
403 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
404 // CHECK1-NEXT:    store i8* null, i8** [[TMP29]], align 8
405 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
406 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
407 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
408 // CHECK1-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
409 // CHECK1-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
410 // CHECK1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
411 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
412 // CHECK1-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
413 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
414 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
415 // CHECK1-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
416 // CHECK1-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
417 // CHECK1-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
418 // CHECK1-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
419 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
420 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
421 // CHECK1-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
422 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
423 // CHECK1-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
424 // CHECK1-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
425 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
426 // CHECK1-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
427 // CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
428 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
429 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
430 // CHECK1-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
431 // CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
432 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
433 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
434 // CHECK1-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
435 // 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)
436 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
437 // CHECK1-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
438 // CHECK1-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
439 // CHECK1-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP38]])
440 // CHECK1-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
441 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
442 // CHECK1-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
443 // CHECK1-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
444 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
445 // CHECK1-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
446 // CHECK1-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
447 // CHECK1-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
448 // CHECK1-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
449 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
450 // CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
451 // CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
452 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
453 // CHECK1-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
454 // CHECK1-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
455 // CHECK1-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
456 // CHECK1-NEXT:    store i8* null, i8** [[TMP65]], align 8
457 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
458 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
459 // CHECK1-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
460 // CHECK1-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
461 // CHECK1-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
462 // CHECK1:       omp_offload.failed:
463 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i64 [[TMP60]]) #[[ATTR3]]
464 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
465 // CHECK1:       omp_offload.cont:
466 // CHECK1-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
467 // CHECK1-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
468 // CHECK1-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
469 // CHECK1-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
470 // CHECK1-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
471 // CHECK1-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
472 // CHECK1-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
473 // CHECK1-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
474 // CHECK1-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
475 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
476 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
477 // CHECK1:       omp_if.then:
478 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
479 // CHECK1-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
480 // CHECK1-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
481 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
482 // CHECK1-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
483 // CHECK1-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
484 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
485 // CHECK1-NEXT:    store i8* null, i8** [[TMP79]], align 8
486 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
487 // CHECK1-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
488 // CHECK1-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
489 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
490 // CHECK1-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
491 // CHECK1-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
492 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
493 // CHECK1-NEXT:    store i8* null, i8** [[TMP84]], align 8
494 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
495 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
496 // CHECK1-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
497 // CHECK1-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
498 // CHECK1-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
499 // CHECK1:       omp_offload.failed19:
500 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
501 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
502 // CHECK1:       omp_offload.cont20:
503 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
504 // CHECK1:       omp_if.else:
505 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
506 // CHECK1-NEXT:    br label [[OMP_IF_END]]
507 // CHECK1:       omp_if.end:
508 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A]], align 4
509 // CHECK1-NEXT:    [[CONV22:%.*]] = bitcast i64* [[A_CASTED21]] to i32*
510 // CHECK1-NEXT:    store i32 [[TMP89]], i32* [[CONV22]], align 4
511 // CHECK1-NEXT:    [[TMP90:%.*]] = load i64, i64* [[A_CASTED21]], align 8
512 // CHECK1-NEXT:    [[TMP91:%.*]] = load i32, i32* [[N_ADDR]], align 4
513 // CHECK1-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP91]], 20
514 // CHECK1-NEXT:    br i1 [[CMP23]], label [[OMP_IF_THEN24:%.*]], label [[OMP_IF_ELSE30:%.*]]
515 // CHECK1:       omp_if.then24:
516 // CHECK1-NEXT:    [[TMP92:%.*]] = mul nuw i64 [[TMP2]], 4
517 // CHECK1-NEXT:    [[TMP93:%.*]] = mul nuw i64 5, [[TMP5]]
518 // CHECK1-NEXT:    [[TMP94:%.*]] = mul nuw i64 [[TMP93]], 8
519 // CHECK1-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
520 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP95]], i8* align 8 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i64 72, i1 false)
521 // CHECK1-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
522 // CHECK1-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i64*
523 // CHECK1-NEXT:    store i64 [[TMP90]], i64* [[TMP97]], align 8
524 // CHECK1-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
525 // CHECK1-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i64*
526 // CHECK1-NEXT:    store i64 [[TMP90]], i64* [[TMP99]], align 8
527 // CHECK1-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 0
528 // CHECK1-NEXT:    store i8* null, i8** [[TMP100]], align 8
529 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 1
530 // CHECK1-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
531 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 8
532 // CHECK1-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 1
533 // CHECK1-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
534 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 8
535 // CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 1
536 // CHECK1-NEXT:    store i8* null, i8** [[TMP105]], align 8
537 // CHECK1-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 2
538 // CHECK1-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
539 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP107]], align 8
540 // CHECK1-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 2
541 // CHECK1-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i64*
542 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP109]], align 8
543 // CHECK1-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 2
544 // CHECK1-NEXT:    store i8* null, i8** [[TMP110]], align 8
545 // CHECK1-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 3
546 // CHECK1-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
547 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP112]], align 8
548 // CHECK1-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 3
549 // CHECK1-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
550 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP114]], align 8
551 // CHECK1-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
552 // CHECK1-NEXT:    store i64 [[TMP92]], i64* [[TMP115]], align 8
553 // CHECK1-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 3
554 // CHECK1-NEXT:    store i8* null, i8** [[TMP116]], align 8
555 // CHECK1-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 4
556 // CHECK1-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
557 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 8
558 // CHECK1-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 4
559 // CHECK1-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
560 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 8
561 // CHECK1-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 4
562 // CHECK1-NEXT:    store i8* null, i8** [[TMP121]], align 8
563 // CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 5
564 // CHECK1-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i64*
565 // CHECK1-NEXT:    store i64 5, i64* [[TMP123]], align 8
566 // CHECK1-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 5
567 // CHECK1-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i64*
568 // CHECK1-NEXT:    store i64 5, i64* [[TMP125]], align 8
569 // CHECK1-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 5
570 // CHECK1-NEXT:    store i8* null, i8** [[TMP126]], align 8
571 // CHECK1-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 6
572 // CHECK1-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
573 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP128]], align 8
574 // CHECK1-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 6
575 // CHECK1-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i64*
576 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP130]], align 8
577 // CHECK1-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 6
578 // CHECK1-NEXT:    store i8* null, i8** [[TMP131]], align 8
579 // CHECK1-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 7
580 // CHECK1-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
581 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 8
582 // CHECK1-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 7
583 // CHECK1-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
584 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 8
585 // CHECK1-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
586 // CHECK1-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 8
587 // CHECK1-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 7
588 // CHECK1-NEXT:    store i8* null, i8** [[TMP137]], align 8
589 // CHECK1-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 8
590 // CHECK1-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
591 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 8
592 // CHECK1-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 8
593 // CHECK1-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
594 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 8
595 // CHECK1-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 8
596 // CHECK1-NEXT:    store i8* null, i8** [[TMP142]], align 8
597 // CHECK1-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
598 // CHECK1-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
599 // CHECK1-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
600 // CHECK1-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
601 // CHECK1-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
602 // CHECK1-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
603 // CHECK1:       omp_offload.failed28:
604 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
605 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
606 // CHECK1:       omp_offload.cont29:
607 // CHECK1-NEXT:    br label [[OMP_IF_END31:%.*]]
608 // CHECK1:       omp_if.else30:
609 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
610 // CHECK1-NEXT:    br label [[OMP_IF_END31]]
611 // CHECK1:       omp_if.end31:
612 // CHECK1-NEXT:    store i32 0, i32* [[NN]], align 4
613 // CHECK1-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
614 // CHECK1-NEXT:    [[CONV32:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
615 // CHECK1-NEXT:    store i32 [[TMP148]], i32* [[CONV32]], align 4
616 // CHECK1-NEXT:    [[TMP149:%.*]] = load i64, i64* [[NN_CASTED]], align 8
617 // CHECK1-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
618 // CHECK1-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
619 // CHECK1-NEXT:    store i64 [[TMP149]], i64* [[TMP151]], align 8
620 // CHECK1-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
621 // CHECK1-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i64*
622 // CHECK1-NEXT:    store i64 [[TMP149]], i64* [[TMP153]], align 8
623 // CHECK1-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i64 0, i64 0
624 // CHECK1-NEXT:    store i8* null, i8** [[TMP154]], align 8
625 // CHECK1-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
626 // CHECK1-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
627 // CHECK1-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
628 // CHECK1-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
629 // CHECK1-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
630 // CHECK1:       omp_offload.failed36:
631 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i64 [[TMP149]]) #[[ATTR3]]
632 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
633 // CHECK1:       omp_offload.cont37:
634 // CHECK1-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
635 // CHECK1-NEXT:    [[CONV39:%.*]] = bitcast i64* [[NN_CASTED38]] to i32*
636 // CHECK1-NEXT:    store i32 [[TMP159]], i32* [[CONV39]], align 4
637 // CHECK1-NEXT:    [[TMP160:%.*]] = load i64, i64* [[NN_CASTED38]], align 8
638 // CHECK1-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
639 // CHECK1-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i64*
640 // CHECK1-NEXT:    store i64 [[TMP160]], i64* [[TMP162]], align 8
641 // CHECK1-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
642 // CHECK1-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i64*
643 // CHECK1-NEXT:    store i64 [[TMP160]], i64* [[TMP164]], align 8
644 // CHECK1-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS42]], i64 0, i64 0
645 // CHECK1-NEXT:    store i8* null, i8** [[TMP165]], align 8
646 // CHECK1-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
647 // CHECK1-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
648 // CHECK1-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
649 // CHECK1-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
650 // CHECK1-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
651 // CHECK1:       omp_offload.failed43:
652 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i64 [[TMP160]]) #[[ATTR3]]
653 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
654 // CHECK1:       omp_offload.cont44:
655 // CHECK1-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
656 // CHECK1-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
657 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
658 // CHECK1-NEXT:    ret i32 [[TMP170]]
659 //
660 //
661 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
662 // CHECK1-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
663 // CHECK1-NEXT:  entry:
664 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
665 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
666 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
667 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
668 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
669 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
670 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
671 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
672 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
673 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
674 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
675 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
676 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
677 // CHECK1-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
678 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
679 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
680 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
681 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
682 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
683 // CHECK1-NEXT:    ret void
684 //
685 //
686 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
687 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
688 // CHECK1-NEXT:  entry:
689 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
690 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
691 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
692 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
693 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
694 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
695 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
696 // CHECK1-NEXT:    ret void
697 //
698 //
699 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
700 // 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]+]] {
701 // CHECK1-NEXT:  entry:
702 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
703 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
704 // CHECK1-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
705 // CHECK1-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
706 // CHECK1-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
707 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
708 // CHECK1-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
709 // CHECK1-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
710 // CHECK1-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
711 // CHECK1-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
712 // CHECK1-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
713 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
714 // CHECK1-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
715 // CHECK1-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
716 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
717 // CHECK1-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
718 // CHECK1-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
719 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
720 // CHECK1-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
721 // CHECK1-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
722 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
723 // CHECK1-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
724 // CHECK1-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
725 // CHECK1-NEXT:    ret void
726 //
727 //
728 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
729 // CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
730 // CHECK1-NEXT:  entry:
731 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
732 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
733 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
734 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
735 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
736 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
737 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
738 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
739 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
740 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
741 // CHECK1-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
742 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
743 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
744 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
745 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
746 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
747 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
748 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
749 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
750 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
751 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
752 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
753 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
754 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
755 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
756 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
757 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
758 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
759 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
760 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
761 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
762 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !24
763 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !24
764 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
765 // 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 !24
766 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !24
767 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
768 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
769 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !24
770 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
771 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
772 // 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]]
773 // CHECK1-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !24
774 // CHECK1-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !24
775 // CHECK1-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !24
776 // CHECK1-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !24
777 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
778 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
779 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
780 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
781 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
782 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
783 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
784 // CHECK1-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
785 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
786 // CHECK1-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
787 // CHECK1:       omp_offload.failed.i:
788 // CHECK1-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
789 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
790 // CHECK1-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !24
791 // CHECK1-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !24
792 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
793 // CHECK1-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
794 // CHECK1-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !24
795 // CHECK1-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !24
796 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
797 // CHECK1-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
798 // CHECK1-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !24
799 // CHECK1-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !24
800 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
801 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
802 // CHECK1:       .omp_outlined..1.exit:
803 // CHECK1-NEXT:    ret i32 0
804 //
805 //
806 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
807 // CHECK1-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
808 // CHECK1-NEXT:  entry:
809 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
810 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
811 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
812 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
813 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
814 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
815 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
816 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
817 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
818 // CHECK1-NEXT:    ret void
819 //
820 //
821 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
822 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
823 // CHECK1-NEXT:  entry:
824 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
825 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
826 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
827 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
828 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
829 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
830 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
831 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
832 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
833 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
834 // CHECK1-NEXT:    ret void
835 //
836 //
837 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
838 // CHECK1-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
839 // CHECK1-NEXT:  entry:
840 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
841 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
842 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
843 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
844 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
845 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
846 // CHECK1-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
847 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
848 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
849 // CHECK1-NEXT:    ret void
850 //
851 //
852 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
853 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
854 // CHECK1-NEXT:  entry:
855 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
856 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
857 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
858 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
859 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
860 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
861 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
862 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
863 // CHECK1-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
864 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
865 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
866 // CHECK1-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
867 // CHECK1-NEXT:    ret void
868 //
869 //
870 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
871 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
872 // CHECK1-NEXT:  entry:
873 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
874 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
875 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
876 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
877 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
878 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
879 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
880 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
881 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
882 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
883 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
884 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
885 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
886 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
887 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
888 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
889 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
890 // CHECK1-NEXT:    ret void
891 //
892 //
893 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
894 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
895 // CHECK1-NEXT:  entry:
896 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
897 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
898 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
899 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
900 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
901 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
902 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
903 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
904 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
905 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
906 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
907 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
908 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
909 // CHECK1-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
910 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
911 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
912 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
913 // CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
914 // CHECK1-NEXT:    ret void
915 //
916 //
917 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
918 // 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:%.*]]) #[[ATTR2]] {
919 // CHECK1-NEXT:  entry:
920 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
921 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
922 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
923 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
924 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
925 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
926 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
927 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
928 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
929 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
930 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
931 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
932 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
933 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
934 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
935 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
936 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
937 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
938 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
939 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
940 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
941 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
942 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
943 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
944 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
945 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
946 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
947 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
948 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
949 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
950 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
951 // CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
952 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
953 // CHECK1-NEXT:    ret void
954 //
955 //
956 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
957 // 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:%.*]]) #[[ATTR2]] {
958 // CHECK1-NEXT:  entry:
959 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
960 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
961 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
962 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
963 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
964 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
965 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
966 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
967 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
968 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
969 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
970 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
971 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
972 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
973 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
974 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
975 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
976 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
977 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
978 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
979 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
980 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
981 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
982 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
983 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
984 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
985 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
986 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
987 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
988 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
989 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
990 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
991 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
992 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
993 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
994 // CHECK1-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
995 // CHECK1-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
996 // CHECK1-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
997 // CHECK1-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
998 // CHECK1-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
999 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
1000 // CHECK1-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
1001 // CHECK1-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
1002 // CHECK1-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
1003 // CHECK1-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
1004 // CHECK1-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
1005 // CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
1006 // CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
1007 // CHECK1-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
1008 // CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
1009 // CHECK1-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
1010 // CHECK1-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
1011 // CHECK1-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
1012 // CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
1013 // CHECK1-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
1014 // CHECK1-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
1015 // CHECK1-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
1016 // CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
1017 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
1018 // CHECK1-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
1019 // CHECK1-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
1020 // CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
1021 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
1022 // CHECK1-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
1023 // CHECK1-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
1024 // CHECK1-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
1025 // CHECK1-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
1026 // CHECK1-NEXT:    ret void
1027 //
1028 //
1029 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
1030 // CHECK1-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
1031 // CHECK1-NEXT:  entry:
1032 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
1033 // CHECK1-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
1034 // CHECK1-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
1035 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
1036 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
1037 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
1038 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
1039 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
1040 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP1]])
1041 // CHECK1-NEXT:    ret void
1042 //
1043 //
1044 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
1045 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
1046 // CHECK1-NEXT:  entry:
1047 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1048 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1049 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
1050 // CHECK1-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
1051 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1052 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1053 // CHECK1-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
1054 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
1055 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
1056 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
1057 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
1058 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
1059 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP1]])
1060 // CHECK1-NEXT:    ret void
1061 //
1062 //
1063 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
1064 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
1065 // CHECK1-NEXT:  entry:
1066 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1067 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1068 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
1069 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1070 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1071 // CHECK1-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
1072 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
1073 // CHECK1-NEXT:    ret void
1074 //
1075 //
1076 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
1077 // CHECK1-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
1078 // CHECK1-NEXT:  entry:
1079 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
1080 // CHECK1-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
1081 // CHECK1-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
1082 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
1083 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
1084 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
1085 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
1086 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
1087 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP1]])
1088 // CHECK1-NEXT:    ret void
1089 //
1090 //
1091 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..16
1092 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
1093 // CHECK1-NEXT:  entry:
1094 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1095 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1096 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
1097 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1098 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1099 // CHECK1-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
1100 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
1101 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[CONV]])
1102 // CHECK1-NEXT:    ret void
1103 //
1104 //
1105 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..17
1106 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
1107 // CHECK1-NEXT:  entry:
1108 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1109 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1110 // CHECK1-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
1111 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1112 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1113 // CHECK1-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
1114 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
1115 // CHECK1-NEXT:    ret void
1116 //
1117 //
1118 // CHECK1-LABEL: define {{[^@]+}}@_Z6bazzzziPi
1119 // CHECK1-SAME: (i32 noundef signext [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
1120 // CHECK1-NEXT:  entry:
1121 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1122 // CHECK1-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 8
1123 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1124 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1125 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1126 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1127 // CHECK1-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 8
1128 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1129 // CHECK1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1130 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1131 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
1132 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
1133 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1134 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
1135 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
1136 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1137 // CHECK1-NEXT:    store i8* null, i8** [[TMP6]], align 8
1138 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1139 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1140 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1141 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
1142 // CHECK1-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1143 // CHECK1:       omp_offload.failed:
1144 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i64 [[TMP1]]) #[[ATTR3]]
1145 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1146 // CHECK1:       omp_offload.cont:
1147 // CHECK1-NEXT:    ret void
1148 //
1149 //
1150 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
1151 // CHECK1-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
1152 // CHECK1-NEXT:  entry:
1153 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1154 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1155 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1156 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i64 [[TMP0]])
1157 // CHECK1-NEXT:    ret void
1158 //
1159 //
1160 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..20
1161 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
1162 // CHECK1-NEXT:  entry:
1163 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1164 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1165 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1166 // CHECK1-NEXT:    [[F:%.*]] = alloca i32*, align 8
1167 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1168 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1169 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1170 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1171 // CHECK1-NEXT:    ret void
1172 //
1173 //
1174 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
1175 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1176 // CHECK1-NEXT:  entry:
1177 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1178 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1179 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
1180 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1181 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1182 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1183 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
1184 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1185 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1186 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1187 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1188 // CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
1189 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1190 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1191 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1192 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1193 // CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
1194 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1195 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1196 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1197 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1198 // CHECK1-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
1199 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1200 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
1201 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
1202 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
1203 // CHECK1-NEXT:    ret i32 [[TMP8]]
1204 //
1205 //
1206 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1207 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1208 // CHECK1-NEXT:  entry:
1209 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1210 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1211 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
1212 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1213 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1214 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1215 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
1216 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
1217 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
1218 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
1219 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1220 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1221 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1222 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1223 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1224 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
1225 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1226 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1227 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1228 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1229 // CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
1230 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
1231 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1232 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1233 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1234 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
1235 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1236 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
1237 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
1238 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1239 // CHECK1:       omp_if.then:
1240 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1241 // CHECK1-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
1242 // CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
1243 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1244 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i64 40, i1 false)
1245 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1246 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
1247 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
1248 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1249 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
1250 // CHECK1-NEXT:    store double* [[A]], double** [[TMP14]], align 8
1251 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1252 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
1253 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1254 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
1255 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
1256 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1257 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
1258 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
1259 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1260 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
1261 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1262 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
1263 // CHECK1-NEXT:    store i64 2, i64* [[TMP22]], align 8
1264 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1265 // CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
1266 // CHECK1-NEXT:    store i64 2, i64* [[TMP24]], align 8
1267 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1268 // CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
1269 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1270 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
1271 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
1272 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1273 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
1274 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
1275 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1276 // CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
1277 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1278 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
1279 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
1280 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1281 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
1282 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
1283 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1284 // CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
1285 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
1286 // CHECK1-NEXT:    store i8* null, i8** [[TMP36]], align 8
1287 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1288 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1289 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1290 // CHECK1-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1291 // CHECK1-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
1292 // CHECK1-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1293 // CHECK1:       omp_offload.failed:
1294 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
1295 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1296 // CHECK1:       omp_offload.cont:
1297 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1298 // CHECK1:       omp_if.else:
1299 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
1300 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1301 // CHECK1:       omp_if.end:
1302 // CHECK1-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
1303 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
1304 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1305 // CHECK1-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
1306 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
1307 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
1308 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
1309 // CHECK1-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1310 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
1311 // CHECK1-NEXT:    ret i32 [[ADD4]]
1312 //
1313 //
1314 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
1315 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1316 // CHECK1-NEXT:  entry:
1317 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1318 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1319 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1320 // CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
1321 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1322 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1323 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1324 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1325 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
1326 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
1327 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
1328 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1329 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1330 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1331 // CHECK1-NEXT:    store i8 0, i8* [[AAA]], align 1
1332 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1333 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1334 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1335 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1336 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1337 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1338 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1339 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1340 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
1341 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1342 // CHECK1-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
1343 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1344 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1345 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
1346 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1347 // CHECK1:       omp_if.then:
1348 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1349 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1350 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1351 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1352 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1353 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
1354 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1355 // CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
1356 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1357 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1358 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1359 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1360 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1361 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
1362 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1363 // CHECK1-NEXT:    store i8* null, i8** [[TMP16]], align 8
1364 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1365 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
1366 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
1367 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1368 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
1369 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
1370 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1371 // CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
1372 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1373 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
1374 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
1375 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1376 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
1377 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
1378 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1379 // CHECK1-NEXT:    store i8* null, i8** [[TMP26]], align 8
1380 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1381 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1382 // CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1383 // CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
1384 // CHECK1-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1385 // CHECK1:       omp_offload.failed:
1386 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
1387 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1388 // CHECK1:       omp_offload.cont:
1389 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1390 // CHECK1:       omp_if.else:
1391 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
1392 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1393 // CHECK1:       omp_if.end:
1394 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
1395 // CHECK1-NEXT:    ret i32 [[TMP31]]
1396 //
1397 //
1398 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1399 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
1400 // CHECK1-NEXT:  entry:
1401 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1402 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1403 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1404 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1405 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1406 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1407 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1408 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1409 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1410 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1411 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1412 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1413 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1414 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1415 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1416 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1417 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1418 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1419 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1420 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1421 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1422 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
1423 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1424 // CHECK1:       omp_if.then:
1425 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1426 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
1427 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
1428 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1429 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1430 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1431 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1432 // CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
1433 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1434 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
1435 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
1436 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1437 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1438 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1439 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1440 // CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
1441 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1442 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
1443 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
1444 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1445 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
1446 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
1447 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1448 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
1449 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1450 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1451 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1452 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
1453 // CHECK1-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1454 // CHECK1:       omp_offload.failed:
1455 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1456 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1457 // CHECK1:       omp_offload.cont:
1458 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1459 // CHECK1:       omp_if.else:
1460 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
1461 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1462 // CHECK1:       omp_if.end:
1463 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
1464 // CHECK1-NEXT:    ret i32 [[TMP24]]
1465 //
1466 //
1467 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
1468 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
1469 // CHECK1-NEXT:  entry:
1470 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1471 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1472 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1473 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1474 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1475 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1476 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1477 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1478 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1479 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1480 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1481 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1482 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1483 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1484 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1485 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1486 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
1487 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1488 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
1489 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
1490 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
1491 // CHECK1-NEXT:    ret void
1492 //
1493 //
1494 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..23
1495 // 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]] {
1496 // CHECK1-NEXT:  entry:
1497 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1498 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1499 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1500 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1501 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1502 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1503 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1504 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1505 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1506 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1507 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1508 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1509 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1510 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1511 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1512 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1513 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1514 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1515 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1516 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
1517 // CHECK1-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
1518 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
1519 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1520 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8
1521 // CHECK1-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
1522 // CHECK1-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
1523 // CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
1524 // CHECK1-NEXT:    store double [[INC]], double* [[A4]], align 8
1525 // CHECK1-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
1526 // CHECK1-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
1527 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
1528 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1529 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
1530 // CHECK1-NEXT:    ret void
1531 //
1532 //
1533 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
1534 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1535 // CHECK1-NEXT:  entry:
1536 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1537 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1538 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1539 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1540 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1541 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1542 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1543 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1544 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1545 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1546 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1547 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1548 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1549 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1550 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1551 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1552 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1553 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
1554 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1555 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
1556 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1557 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
1558 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1559 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
1560 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1561 // CHECK1-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
1562 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1563 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
1564 // CHECK1-NEXT:    ret void
1565 //
1566 //
1567 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..26
1568 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1569 // CHECK1-NEXT:  entry:
1570 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1571 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1572 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1573 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1574 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1575 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1576 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1577 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1578 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1579 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1580 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1581 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1582 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1583 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1584 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1585 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1586 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1587 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
1588 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
1589 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
1590 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
1591 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
1592 // CHECK1-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
1593 // CHECK1-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
1594 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
1595 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
1596 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
1597 // CHECK1-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
1598 // CHECK1-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
1599 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
1600 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1601 // CHECK1-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
1602 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
1603 // CHECK1-NEXT:    ret void
1604 //
1605 //
1606 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
1607 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1608 // CHECK1-NEXT:  entry:
1609 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1610 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1611 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1612 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1613 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1614 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1615 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1616 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1617 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1618 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1619 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1620 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1621 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1622 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
1623 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1624 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
1625 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1626 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
1627 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1628 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
1629 // CHECK1-NEXT:    ret void
1630 //
1631 //
1632 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..29
1633 // 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]] {
1634 // CHECK1-NEXT:  entry:
1635 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1636 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1637 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1638 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1639 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1640 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1641 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1642 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1643 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1644 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1645 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1646 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1647 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1648 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1649 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
1650 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
1651 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
1652 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
1653 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
1654 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
1655 // CHECK1-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
1656 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
1657 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1658 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
1659 // CHECK1-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
1660 // CHECK1-NEXT:    ret void
1661 //
1662 //
1663 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1664 // CHECK1-SAME: () #[[ATTR4]] {
1665 // CHECK1-NEXT:  entry:
1666 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1667 // CHECK1-NEXT:    ret void
1668 //
1669 //
1670 // CHECK2-LABEL: define {{[^@]+}}@_Z3fooi
1671 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1672 // CHECK2-NEXT:  entry:
1673 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1674 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
1675 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
1676 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
1677 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1678 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1679 // CHECK2-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
1680 // CHECK2-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
1681 // CHECK2-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
1682 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1683 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1684 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1685 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1686 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
1687 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1688 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1689 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1690 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
1691 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1692 // CHECK2-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
1693 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
1694 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
1695 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
1696 // CHECK2-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
1697 // CHECK2-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
1698 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
1699 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
1700 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
1701 // CHECK2-NEXT:    [[A_CASTED21:%.*]] = alloca i64, align 8
1702 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS25:%.*]] = alloca [9 x i8*], align 8
1703 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS26:%.*]] = alloca [9 x i8*], align 8
1704 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS27:%.*]] = alloca [9 x i8*], align 8
1705 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
1706 // CHECK2-NEXT:    [[NN:%.*]] = alloca i32, align 4
1707 // CHECK2-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
1708 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 8
1709 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 8
1710 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 8
1711 // CHECK2-NEXT:    [[NN_CASTED38:%.*]] = alloca i64, align 8
1712 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS40:%.*]] = alloca [1 x i8*], align 8
1713 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS41:%.*]] = alloca [1 x i8*], align 8
1714 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS42:%.*]] = alloca [1 x i8*], align 8
1715 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
1716 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1717 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
1718 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
1719 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1720 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1721 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1722 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1723 // CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
1724 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1725 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1726 // CHECK2-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
1727 // CHECK2-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
1728 // CHECK2-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
1729 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
1730 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1731 // CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
1732 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
1733 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
1734 // CHECK2-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
1735 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1736 // CHECK2-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
1737 // CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1738 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1739 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
1740 // CHECK2-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
1741 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
1742 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1743 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
1744 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
1745 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
1746 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1747 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
1748 // CHECK2-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
1749 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1750 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
1751 // CHECK2-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
1752 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1753 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
1754 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1755 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
1756 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
1757 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1758 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
1759 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
1760 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1761 // CHECK2-NEXT:    store i8* null, i8** [[TMP24]], align 8
1762 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1763 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
1764 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
1765 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1766 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
1767 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
1768 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1769 // CHECK2-NEXT:    store i8* null, i8** [[TMP29]], align 8
1770 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1771 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1772 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
1773 // CHECK2-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
1774 // CHECK2-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
1775 // CHECK2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
1776 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1777 // CHECK2-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
1778 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
1779 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1780 // CHECK2-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
1781 // CHECK2-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
1782 // CHECK2-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
1783 // CHECK2-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
1784 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
1785 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
1786 // CHECK2-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
1787 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
1788 // CHECK2-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
1789 // CHECK2-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
1790 // CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
1791 // CHECK2-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
1792 // CHECK2-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
1793 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
1794 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
1795 // CHECK2-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
1796 // CHECK2-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
1797 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
1798 // CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
1799 // CHECK2-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
1800 // 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)
1801 // CHECK2-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
1802 // CHECK2-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
1803 // CHECK2-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
1804 // CHECK2-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP38]])
1805 // CHECK2-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
1806 // CHECK2-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1807 // CHECK2-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
1808 // CHECK2-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
1809 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
1810 // CHECK2-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
1811 // CHECK2-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
1812 // CHECK2-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
1813 // CHECK2-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
1814 // CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
1815 // CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
1816 // CHECK2-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
1817 // CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
1818 // CHECK2-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
1819 // CHECK2-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
1820 // CHECK2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
1821 // CHECK2-NEXT:    store i8* null, i8** [[TMP65]], align 8
1822 // CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
1823 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
1824 // CHECK2-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
1825 // CHECK2-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
1826 // CHECK2-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1827 // CHECK2:       omp_offload.failed:
1828 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i64 [[TMP60]]) #[[ATTR3]]
1829 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1830 // CHECK2:       omp_offload.cont:
1831 // CHECK2-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
1832 // CHECK2-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
1833 // CHECK2-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
1834 // CHECK2-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
1835 // CHECK2-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
1836 // CHECK2-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
1837 // CHECK2-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
1838 // CHECK2-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
1839 // CHECK2-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
1840 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
1841 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1842 // CHECK2:       omp_if.then:
1843 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
1844 // CHECK2-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
1845 // CHECK2-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
1846 // CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
1847 // CHECK2-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
1848 // CHECK2-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
1849 // CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
1850 // CHECK2-NEXT:    store i8* null, i8** [[TMP79]], align 8
1851 // CHECK2-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
1852 // CHECK2-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
1853 // CHECK2-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
1854 // CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
1855 // CHECK2-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
1856 // CHECK2-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
1857 // CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
1858 // CHECK2-NEXT:    store i8* null, i8** [[TMP84]], align 8
1859 // CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
1860 // CHECK2-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
1861 // CHECK2-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
1862 // CHECK2-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
1863 // CHECK2-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
1864 // CHECK2:       omp_offload.failed19:
1865 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
1866 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
1867 // CHECK2:       omp_offload.cont20:
1868 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
1869 // CHECK2:       omp_if.else:
1870 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
1871 // CHECK2-NEXT:    br label [[OMP_IF_END]]
1872 // CHECK2:       omp_if.end:
1873 // CHECK2-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A]], align 4
1874 // CHECK2-NEXT:    [[CONV22:%.*]] = bitcast i64* [[A_CASTED21]] to i32*
1875 // CHECK2-NEXT:    store i32 [[TMP89]], i32* [[CONV22]], align 4
1876 // CHECK2-NEXT:    [[TMP90:%.*]] = load i64, i64* [[A_CASTED21]], align 8
1877 // CHECK2-NEXT:    [[TMP91:%.*]] = load i32, i32* [[N_ADDR]], align 4
1878 // CHECK2-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP91]], 20
1879 // CHECK2-NEXT:    br i1 [[CMP23]], label [[OMP_IF_THEN24:%.*]], label [[OMP_IF_ELSE30:%.*]]
1880 // CHECK2:       omp_if.then24:
1881 // CHECK2-NEXT:    [[TMP92:%.*]] = mul nuw i64 [[TMP2]], 4
1882 // CHECK2-NEXT:    [[TMP93:%.*]] = mul nuw i64 5, [[TMP5]]
1883 // CHECK2-NEXT:    [[TMP94:%.*]] = mul nuw i64 [[TMP93]], 8
1884 // CHECK2-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1885 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP95]], i8* align 8 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i64 72, i1 false)
1886 // CHECK2-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
1887 // CHECK2-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i64*
1888 // CHECK2-NEXT:    store i64 [[TMP90]], i64* [[TMP97]], align 8
1889 // CHECK2-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
1890 // CHECK2-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i64*
1891 // CHECK2-NEXT:    store i64 [[TMP90]], i64* [[TMP99]], align 8
1892 // CHECK2-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 0
1893 // CHECK2-NEXT:    store i8* null, i8** [[TMP100]], align 8
1894 // CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 1
1895 // CHECK2-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
1896 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 8
1897 // CHECK2-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 1
1898 // CHECK2-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
1899 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 8
1900 // CHECK2-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 1
1901 // CHECK2-NEXT:    store i8* null, i8** [[TMP105]], align 8
1902 // CHECK2-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 2
1903 // CHECK2-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
1904 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP107]], align 8
1905 // CHECK2-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 2
1906 // CHECK2-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i64*
1907 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP109]], align 8
1908 // CHECK2-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 2
1909 // CHECK2-NEXT:    store i8* null, i8** [[TMP110]], align 8
1910 // CHECK2-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 3
1911 // CHECK2-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
1912 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP112]], align 8
1913 // CHECK2-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 3
1914 // CHECK2-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
1915 // CHECK2-NEXT:    store float* [[VLA]], float** [[TMP114]], align 8
1916 // CHECK2-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
1917 // CHECK2-NEXT:    store i64 [[TMP92]], i64* [[TMP115]], align 8
1918 // CHECK2-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 3
1919 // CHECK2-NEXT:    store i8* null, i8** [[TMP116]], align 8
1920 // CHECK2-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 4
1921 // CHECK2-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
1922 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 8
1923 // CHECK2-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 4
1924 // CHECK2-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
1925 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 8
1926 // CHECK2-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 4
1927 // CHECK2-NEXT:    store i8* null, i8** [[TMP121]], align 8
1928 // CHECK2-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 5
1929 // CHECK2-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i64*
1930 // CHECK2-NEXT:    store i64 5, i64* [[TMP123]], align 8
1931 // CHECK2-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 5
1932 // CHECK2-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i64*
1933 // CHECK2-NEXT:    store i64 5, i64* [[TMP125]], align 8
1934 // CHECK2-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 5
1935 // CHECK2-NEXT:    store i8* null, i8** [[TMP126]], align 8
1936 // CHECK2-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 6
1937 // CHECK2-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
1938 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP128]], align 8
1939 // CHECK2-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 6
1940 // CHECK2-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i64*
1941 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP130]], align 8
1942 // CHECK2-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 6
1943 // CHECK2-NEXT:    store i8* null, i8** [[TMP131]], align 8
1944 // CHECK2-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 7
1945 // CHECK2-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
1946 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 8
1947 // CHECK2-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 7
1948 // CHECK2-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
1949 // CHECK2-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 8
1950 // CHECK2-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
1951 // CHECK2-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 8
1952 // CHECK2-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 7
1953 // CHECK2-NEXT:    store i8* null, i8** [[TMP137]], align 8
1954 // CHECK2-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 8
1955 // CHECK2-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
1956 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 8
1957 // CHECK2-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 8
1958 // CHECK2-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
1959 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 8
1960 // CHECK2-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 8
1961 // CHECK2-NEXT:    store i8* null, i8** [[TMP142]], align 8
1962 // CHECK2-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
1963 // CHECK2-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
1964 // CHECK2-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1965 // CHECK2-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1966 // CHECK2-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
1967 // CHECK2-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
1968 // CHECK2:       omp_offload.failed28:
1969 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
1970 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
1971 // CHECK2:       omp_offload.cont29:
1972 // CHECK2-NEXT:    br label [[OMP_IF_END31:%.*]]
1973 // CHECK2:       omp_if.else30:
1974 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
1975 // CHECK2-NEXT:    br label [[OMP_IF_END31]]
1976 // CHECK2:       omp_if.end31:
1977 // CHECK2-NEXT:    store i32 0, i32* [[NN]], align 4
1978 // CHECK2-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
1979 // CHECK2-NEXT:    [[CONV32:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
1980 // CHECK2-NEXT:    store i32 [[TMP148]], i32* [[CONV32]], align 4
1981 // CHECK2-NEXT:    [[TMP149:%.*]] = load i64, i64* [[NN_CASTED]], align 8
1982 // CHECK2-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
1983 // CHECK2-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
1984 // CHECK2-NEXT:    store i64 [[TMP149]], i64* [[TMP151]], align 8
1985 // CHECK2-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
1986 // CHECK2-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i64*
1987 // CHECK2-NEXT:    store i64 [[TMP149]], i64* [[TMP153]], align 8
1988 // CHECK2-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i64 0, i64 0
1989 // CHECK2-NEXT:    store i8* null, i8** [[TMP154]], align 8
1990 // CHECK2-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
1991 // CHECK2-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
1992 // CHECK2-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1993 // CHECK2-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
1994 // CHECK2-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
1995 // CHECK2:       omp_offload.failed36:
1996 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i64 [[TMP149]]) #[[ATTR3]]
1997 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
1998 // CHECK2:       omp_offload.cont37:
1999 // CHECK2-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
2000 // CHECK2-NEXT:    [[CONV39:%.*]] = bitcast i64* [[NN_CASTED38]] to i32*
2001 // CHECK2-NEXT:    store i32 [[TMP159]], i32* [[CONV39]], align 4
2002 // CHECK2-NEXT:    [[TMP160:%.*]] = load i64, i64* [[NN_CASTED38]], align 8
2003 // CHECK2-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
2004 // CHECK2-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i64*
2005 // CHECK2-NEXT:    store i64 [[TMP160]], i64* [[TMP162]], align 8
2006 // CHECK2-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
2007 // CHECK2-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i64*
2008 // CHECK2-NEXT:    store i64 [[TMP160]], i64* [[TMP164]], align 8
2009 // CHECK2-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS42]], i64 0, i64 0
2010 // CHECK2-NEXT:    store i8* null, i8** [[TMP165]], align 8
2011 // CHECK2-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
2012 // CHECK2-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
2013 // CHECK2-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2014 // CHECK2-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
2015 // CHECK2-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
2016 // CHECK2:       omp_offload.failed43:
2017 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i64 [[TMP160]]) #[[ATTR3]]
2018 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
2019 // CHECK2:       omp_offload.cont44:
2020 // CHECK2-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
2021 // CHECK2-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2022 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
2023 // CHECK2-NEXT:    ret i32 [[TMP170]]
2024 //
2025 //
2026 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
2027 // CHECK2-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
2028 // CHECK2-NEXT:  entry:
2029 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2030 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2031 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
2032 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2033 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
2034 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2035 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2036 // CHECK2-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
2037 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2038 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2039 // CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
2040 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
2041 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
2042 // CHECK2-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
2043 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
2044 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2045 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
2046 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2047 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
2048 // CHECK2-NEXT:    ret void
2049 //
2050 //
2051 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
2052 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2053 // CHECK2-NEXT:  entry:
2054 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2055 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2056 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2057 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2058 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2059 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2060 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2061 // CHECK2-NEXT:    ret void
2062 //
2063 //
2064 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
2065 // 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]+]] {
2066 // CHECK2-NEXT:  entry:
2067 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
2068 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
2069 // CHECK2-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
2070 // CHECK2-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
2071 // CHECK2-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
2072 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
2073 // CHECK2-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
2074 // CHECK2-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
2075 // CHECK2-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
2076 // CHECK2-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
2077 // CHECK2-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
2078 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
2079 // CHECK2-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
2080 // CHECK2-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
2081 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
2082 // CHECK2-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
2083 // CHECK2-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
2084 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
2085 // CHECK2-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
2086 // CHECK2-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
2087 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
2088 // CHECK2-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
2089 // CHECK2-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
2090 // CHECK2-NEXT:    ret void
2091 //
2092 //
2093 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
2094 // CHECK2-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
2095 // CHECK2-NEXT:  entry:
2096 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2097 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
2098 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
2099 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
2100 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
2101 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
2102 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
2103 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
2104 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
2105 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
2106 // CHECK2-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
2107 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
2108 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
2109 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2110 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
2111 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2112 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2113 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2114 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
2115 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2116 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2117 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2118 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
2119 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2120 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
2121 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
2122 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2123 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
2124 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
2125 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
2126 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
2127 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !24
2128 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !24
2129 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
2130 // 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 !24
2131 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !24
2132 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
2133 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
2134 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !24
2135 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
2136 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
2137 // 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]]
2138 // CHECK2-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !24
2139 // CHECK2-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !24
2140 // CHECK2-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !24
2141 // CHECK2-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !24
2142 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
2143 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
2144 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
2145 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
2146 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
2147 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
2148 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
2149 // CHECK2-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
2150 // CHECK2-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
2151 // CHECK2-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2152 // CHECK2:       omp_offload.failed.i:
2153 // CHECK2-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
2154 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
2155 // CHECK2-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !24
2156 // CHECK2-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !24
2157 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
2158 // CHECK2-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
2159 // CHECK2-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !24
2160 // CHECK2-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !24
2161 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
2162 // CHECK2-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
2163 // CHECK2-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !24
2164 // CHECK2-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !24
2165 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
2166 // CHECK2-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2167 // CHECK2:       .omp_outlined..1.exit:
2168 // CHECK2-NEXT:    ret i32 0
2169 //
2170 //
2171 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
2172 // CHECK2-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
2173 // CHECK2-NEXT:  entry:
2174 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2175 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2176 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2177 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2178 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2179 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2180 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
2181 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2182 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2183 // CHECK2-NEXT:    ret void
2184 //
2185 //
2186 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
2187 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
2188 // CHECK2-NEXT:  entry:
2189 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2190 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2191 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2192 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2193 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2194 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2195 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2196 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2197 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
2198 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
2199 // CHECK2-NEXT:    ret void
2200 //
2201 //
2202 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
2203 // CHECK2-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2204 // CHECK2-NEXT:  entry:
2205 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2206 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2207 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2208 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2209 // CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
2210 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2211 // CHECK2-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
2212 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2213 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2214 // CHECK2-NEXT:    ret void
2215 //
2216 //
2217 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
2218 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2219 // CHECK2-NEXT:  entry:
2220 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2221 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2222 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2223 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2224 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2225 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2226 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2227 // CHECK2-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
2228 // CHECK2-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
2229 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
2230 // CHECK2-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
2231 // CHECK2-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
2232 // CHECK2-NEXT:    ret void
2233 //
2234 //
2235 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
2236 // CHECK2-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2237 // CHECK2-NEXT:  entry:
2238 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2239 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2240 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2241 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2242 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2243 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2244 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2245 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2246 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2247 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2248 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
2249 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2250 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
2251 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2252 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
2253 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2254 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
2255 // CHECK2-NEXT:    ret void
2256 //
2257 //
2258 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
2259 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
2260 // CHECK2-NEXT:  entry:
2261 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2262 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2263 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2264 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2265 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2266 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2267 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2268 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2269 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2270 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2271 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2272 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
2273 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
2274 // CHECK2-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
2275 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
2276 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
2277 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
2278 // CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
2279 // CHECK2-NEXT:    ret void
2280 //
2281 //
2282 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
2283 // 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:%.*]]) #[[ATTR2]] {
2284 // CHECK2-NEXT:  entry:
2285 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2286 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
2287 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2288 // CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
2289 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
2290 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2291 // CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
2292 // CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
2293 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
2294 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2295 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2296 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
2297 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2298 // CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
2299 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
2300 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2301 // CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
2302 // CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
2303 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
2304 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2305 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
2306 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2307 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
2308 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
2309 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2310 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
2311 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
2312 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
2313 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
2314 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2315 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
2316 // CHECK2-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
2317 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
2318 // CHECK2-NEXT:    ret void
2319 //
2320 //
2321 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
2322 // 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:%.*]]) #[[ATTR2]] {
2323 // CHECK2-NEXT:  entry:
2324 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2325 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2326 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2327 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
2328 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2329 // CHECK2-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
2330 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
2331 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2332 // CHECK2-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
2333 // CHECK2-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
2334 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
2335 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2336 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2337 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2338 // CHECK2-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
2339 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2340 // CHECK2-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
2341 // CHECK2-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
2342 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2343 // CHECK2-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
2344 // CHECK2-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
2345 // CHECK2-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
2346 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2347 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
2348 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2349 // CHECK2-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
2350 // CHECK2-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
2351 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2352 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
2353 // CHECK2-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
2354 // CHECK2-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
2355 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
2356 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
2357 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
2358 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
2359 // CHECK2-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
2360 // CHECK2-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
2361 // CHECK2-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
2362 // CHECK2-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
2363 // CHECK2-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
2364 // CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
2365 // CHECK2-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
2366 // CHECK2-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
2367 // CHECK2-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
2368 // CHECK2-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
2369 // CHECK2-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
2370 // CHECK2-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
2371 // CHECK2-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
2372 // CHECK2-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
2373 // CHECK2-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
2374 // CHECK2-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
2375 // CHECK2-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
2376 // CHECK2-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
2377 // CHECK2-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
2378 // CHECK2-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
2379 // CHECK2-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
2380 // CHECK2-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
2381 // CHECK2-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
2382 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
2383 // CHECK2-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
2384 // CHECK2-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
2385 // CHECK2-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
2386 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
2387 // CHECK2-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
2388 // CHECK2-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
2389 // CHECK2-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
2390 // CHECK2-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
2391 // CHECK2-NEXT:    ret void
2392 //
2393 //
2394 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
2395 // CHECK2-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
2396 // CHECK2-NEXT:  entry:
2397 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
2398 // CHECK2-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
2399 // CHECK2-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
2400 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
2401 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2402 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
2403 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
2404 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
2405 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2406 // CHECK2-NEXT:    ret void
2407 //
2408 //
2409 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..12
2410 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
2411 // CHECK2-NEXT:  entry:
2412 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2413 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2414 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
2415 // CHECK2-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
2416 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2417 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2418 // CHECK2-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
2419 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
2420 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2421 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
2422 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
2423 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
2424 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2425 // CHECK2-NEXT:    ret void
2426 //
2427 //
2428 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..13
2429 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
2430 // CHECK2-NEXT:  entry:
2431 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2432 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2433 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
2434 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2435 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2436 // CHECK2-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
2437 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
2438 // CHECK2-NEXT:    ret void
2439 //
2440 //
2441 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
2442 // CHECK2-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
2443 // CHECK2-NEXT:  entry:
2444 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
2445 // CHECK2-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
2446 // CHECK2-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
2447 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
2448 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
2449 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
2450 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
2451 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
2452 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP1]])
2453 // CHECK2-NEXT:    ret void
2454 //
2455 //
2456 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..16
2457 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
2458 // CHECK2-NEXT:  entry:
2459 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2460 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2461 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
2462 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2463 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2464 // CHECK2-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
2465 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
2466 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[CONV]])
2467 // CHECK2-NEXT:    ret void
2468 //
2469 //
2470 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..17
2471 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
2472 // CHECK2-NEXT:  entry:
2473 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2474 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2475 // CHECK2-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
2476 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2477 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2478 // CHECK2-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
2479 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
2480 // CHECK2-NEXT:    ret void
2481 //
2482 //
2483 // CHECK2-LABEL: define {{[^@]+}}@_Z6bazzzziPi
2484 // CHECK2-SAME: (i32 noundef signext [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
2485 // CHECK2-NEXT:  entry:
2486 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2487 // CHECK2-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 8
2488 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2489 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2490 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2491 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2492 // CHECK2-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 8
2493 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2494 // CHECK2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
2495 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2496 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
2497 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
2498 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2499 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
2500 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
2501 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2502 // CHECK2-NEXT:    store i8* null, i8** [[TMP6]], align 8
2503 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2504 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2505 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2506 // CHECK2-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
2507 // CHECK2-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2508 // CHECK2:       omp_offload.failed:
2509 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i64 [[TMP1]]) #[[ATTR3]]
2510 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2511 // CHECK2:       omp_offload.cont:
2512 // CHECK2-NEXT:    ret void
2513 //
2514 //
2515 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
2516 // CHECK2-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
2517 // CHECK2-NEXT:  entry:
2518 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2519 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2520 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2521 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i64 [[TMP0]])
2522 // CHECK2-NEXT:    ret void
2523 //
2524 //
2525 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..20
2526 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
2527 // CHECK2-NEXT:  entry:
2528 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2529 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2530 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2531 // CHECK2-NEXT:    [[F:%.*]] = alloca i32*, align 8
2532 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2533 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2534 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2535 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2536 // CHECK2-NEXT:    ret void
2537 //
2538 //
2539 // CHECK2-LABEL: define {{[^@]+}}@_Z3bari
2540 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
2541 // CHECK2-NEXT:  entry:
2542 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2543 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2544 // CHECK2-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
2545 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2546 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2547 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2548 // CHECK2-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
2549 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
2550 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
2551 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2552 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2553 // CHECK2-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
2554 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2555 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
2556 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
2557 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2558 // CHECK2-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
2559 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2560 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
2561 // CHECK2-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
2562 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2563 // CHECK2-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
2564 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2565 // CHECK2-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
2566 // CHECK2-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
2567 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
2568 // CHECK2-NEXT:    ret i32 [[TMP8]]
2569 //
2570 //
2571 // CHECK2-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
2572 // CHECK2-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
2573 // CHECK2-NEXT:  entry:
2574 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2575 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2576 // CHECK2-NEXT:    [[B:%.*]] = alloca i32, align 4
2577 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
2578 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2579 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
2580 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
2581 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
2582 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
2583 // CHECK2-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
2584 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2585 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2586 // CHECK2-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2587 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2588 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
2589 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
2590 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
2591 // CHECK2-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
2592 // CHECK2-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
2593 // CHECK2-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
2594 // CHECK2-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
2595 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
2596 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
2597 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
2598 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
2599 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
2600 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
2601 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
2602 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
2603 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2604 // CHECK2:       omp_if.then:
2605 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
2606 // CHECK2-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
2607 // CHECK2-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
2608 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2609 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i64 40, i1 false)
2610 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2611 // CHECK2-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
2612 // CHECK2-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
2613 // CHECK2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2614 // CHECK2-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
2615 // CHECK2-NEXT:    store double* [[A]], double** [[TMP14]], align 8
2616 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2617 // CHECK2-NEXT:    store i8* null, i8** [[TMP15]], align 8
2618 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2619 // CHECK2-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
2620 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
2621 // CHECK2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2622 // CHECK2-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
2623 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
2624 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2625 // CHECK2-NEXT:    store i8* null, i8** [[TMP20]], align 8
2626 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2627 // CHECK2-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
2628 // CHECK2-NEXT:    store i64 2, i64* [[TMP22]], align 8
2629 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2630 // CHECK2-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
2631 // CHECK2-NEXT:    store i64 2, i64* [[TMP24]], align 8
2632 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2633 // CHECK2-NEXT:    store i8* null, i8** [[TMP25]], align 8
2634 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2635 // CHECK2-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
2636 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
2637 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2638 // CHECK2-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
2639 // CHECK2-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
2640 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
2641 // CHECK2-NEXT:    store i8* null, i8** [[TMP30]], align 8
2642 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
2643 // CHECK2-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
2644 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
2645 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
2646 // CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
2647 // CHECK2-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
2648 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
2649 // CHECK2-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
2650 // CHECK2-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
2651 // CHECK2-NEXT:    store i8* null, i8** [[TMP36]], align 8
2652 // CHECK2-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2653 // CHECK2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2654 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2655 // CHECK2-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2656 // CHECK2-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
2657 // CHECK2-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2658 // CHECK2:       omp_offload.failed:
2659 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
2660 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2661 // CHECK2:       omp_offload.cont:
2662 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2663 // CHECK2:       omp_if.else:
2664 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
2665 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2666 // CHECK2:       omp_if.end:
2667 // CHECK2-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
2668 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
2669 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
2670 // CHECK2-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
2671 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
2672 // CHECK2-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
2673 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
2674 // CHECK2-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2675 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
2676 // CHECK2-NEXT:    ret i32 [[ADD4]]
2677 //
2678 //
2679 // CHECK2-LABEL: define {{[^@]+}}@_ZL7fstatici
2680 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
2681 // CHECK2-NEXT:  entry:
2682 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2683 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2684 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2685 // CHECK2-NEXT:    [[AAA:%.*]] = alloca i8, align 1
2686 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
2687 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2688 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2689 // CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
2690 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
2691 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
2692 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
2693 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2694 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2695 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
2696 // CHECK2-NEXT:    store i8 0, i8* [[AAA]], align 1
2697 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
2698 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2699 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
2700 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2701 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
2702 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2703 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
2704 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2705 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
2706 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
2707 // CHECK2-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
2708 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
2709 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2710 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
2711 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2712 // CHECK2:       omp_if.then:
2713 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2714 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
2715 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
2716 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2717 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
2718 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
2719 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2720 // CHECK2-NEXT:    store i8* null, i8** [[TMP11]], align 8
2721 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2722 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
2723 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
2724 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2725 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
2726 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
2727 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2728 // CHECK2-NEXT:    store i8* null, i8** [[TMP16]], align 8
2729 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2730 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
2731 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
2732 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2733 // CHECK2-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
2734 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
2735 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2736 // CHECK2-NEXT:    store i8* null, i8** [[TMP21]], align 8
2737 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2738 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
2739 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
2740 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2741 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
2742 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
2743 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
2744 // CHECK2-NEXT:    store i8* null, i8** [[TMP26]], align 8
2745 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2746 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2747 // CHECK2-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2748 // CHECK2-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
2749 // CHECK2-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2750 // CHECK2:       omp_offload.failed:
2751 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
2752 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2753 // CHECK2:       omp_offload.cont:
2754 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2755 // CHECK2:       omp_if.else:
2756 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
2757 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2758 // CHECK2:       omp_if.end:
2759 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
2760 // CHECK2-NEXT:    ret i32 [[TMP31]]
2761 //
2762 //
2763 // CHECK2-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
2764 // CHECK2-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
2765 // CHECK2-NEXT:  entry:
2766 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2767 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
2768 // CHECK2-NEXT:    [[AA:%.*]] = alloca i16, align 2
2769 // CHECK2-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
2770 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2771 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2772 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2773 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2774 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2775 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2776 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
2777 // CHECK2-NEXT:    store i16 0, i16* [[AA]], align 2
2778 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
2779 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2780 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
2781 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
2782 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
2783 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2784 // CHECK2-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
2785 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2786 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2787 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
2788 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2789 // CHECK2:       omp_if.then:
2790 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2791 // CHECK2-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
2792 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
2793 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2794 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
2795 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
2796 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2797 // CHECK2-NEXT:    store i8* null, i8** [[TMP9]], align 8
2798 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2799 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
2800 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
2801 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2802 // CHECK2-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
2803 // CHECK2-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
2804 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2805 // CHECK2-NEXT:    store i8* null, i8** [[TMP14]], align 8
2806 // CHECK2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2807 // CHECK2-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
2808 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
2809 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2810 // CHECK2-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
2811 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
2812 // CHECK2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2813 // CHECK2-NEXT:    store i8* null, i8** [[TMP19]], align 8
2814 // CHECK2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2815 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2816 // CHECK2-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2817 // CHECK2-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
2818 // CHECK2-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2819 // CHECK2:       omp_offload.failed:
2820 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
2821 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2822 // CHECK2:       omp_offload.cont:
2823 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2824 // CHECK2:       omp_if.else:
2825 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
2826 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2827 // CHECK2:       omp_if.end:
2828 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
2829 // CHECK2-NEXT:    ret i32 [[TMP24]]
2830 //
2831 //
2832 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
2833 // CHECK2-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
2834 // CHECK2-NEXT:  entry:
2835 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2836 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2837 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2838 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2839 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
2840 // CHECK2-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
2841 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2842 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2843 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2844 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2845 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
2846 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2847 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
2848 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2849 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2850 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
2851 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
2852 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
2853 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
2854 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
2855 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
2856 // CHECK2-NEXT:    ret void
2857 //
2858 //
2859 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..23
2860 // 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]] {
2861 // CHECK2-NEXT:  entry:
2862 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2863 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2864 // CHECK2-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
2865 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
2866 // CHECK2-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
2867 // CHECK2-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
2868 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
2869 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2870 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2871 // CHECK2-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
2872 // CHECK2-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
2873 // CHECK2-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2874 // CHECK2-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
2875 // CHECK2-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
2876 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
2877 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
2878 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2879 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
2880 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
2881 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
2882 // CHECK2-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
2883 // CHECK2-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
2884 // CHECK2-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
2885 // CHECK2-NEXT:    store double [[ADD]], double* [[A]], align 8
2886 // CHECK2-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
2887 // CHECK2-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
2888 // CHECK2-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
2889 // CHECK2-NEXT:    store double [[INC]], double* [[A4]], align 8
2890 // CHECK2-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
2891 // CHECK2-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
2892 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
2893 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
2894 // CHECK2-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
2895 // CHECK2-NEXT:    ret void
2896 //
2897 //
2898 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
2899 // CHECK2-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
2900 // CHECK2-NEXT:  entry:
2901 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2902 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2903 // CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
2904 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
2905 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2906 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2907 // CHECK2-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
2908 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2909 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2910 // CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
2911 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
2912 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2913 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2914 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
2915 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
2916 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
2917 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2918 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
2919 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
2920 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
2921 // CHECK2-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2922 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
2923 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2924 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
2925 // CHECK2-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
2926 // CHECK2-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
2927 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
2928 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
2929 // CHECK2-NEXT:    ret void
2930 //
2931 //
2932 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..26
2933 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
2934 // CHECK2-NEXT:  entry:
2935 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2936 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2937 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2938 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2939 // CHECK2-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
2940 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
2941 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2942 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2943 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2944 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2945 // CHECK2-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
2946 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
2947 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2948 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2949 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
2950 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
2951 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
2952 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
2953 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
2954 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
2955 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
2956 // CHECK2-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
2957 // CHECK2-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
2958 // CHECK2-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
2959 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
2960 // CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
2961 // CHECK2-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
2962 // CHECK2-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
2963 // CHECK2-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
2964 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
2965 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2966 // CHECK2-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
2967 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
2968 // CHECK2-NEXT:    ret void
2969 //
2970 //
2971 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
2972 // CHECK2-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
2973 // CHECK2-NEXT:  entry:
2974 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
2975 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
2976 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
2977 // CHECK2-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
2978 // CHECK2-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
2979 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
2980 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
2981 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
2982 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
2983 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
2984 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
2985 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
2986 // CHECK2-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
2987 // CHECK2-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
2988 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
2989 // CHECK2-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
2990 // CHECK2-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
2991 // CHECK2-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
2992 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
2993 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
2994 // CHECK2-NEXT:    ret void
2995 //
2996 //
2997 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..29
2998 // 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]] {
2999 // CHECK2-NEXT:  entry:
3000 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3001 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3002 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3003 // CHECK2-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3004 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
3005 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3006 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3007 // CHECK2-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3008 // CHECK2-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3009 // CHECK2-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
3010 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3011 // CHECK2-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3012 // CHECK2-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
3013 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
3014 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
3015 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
3016 // CHECK2-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
3017 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
3018 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
3019 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
3020 // CHECK2-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
3021 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
3022 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3023 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
3024 // CHECK2-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
3025 // CHECK2-NEXT:    ret void
3026 //
3027 //
3028 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3029 // CHECK2-SAME: () #[[ATTR4]] {
3030 // CHECK2-NEXT:  entry:
3031 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
3032 // CHECK2-NEXT:    ret void
3033 //
3034 //
3035 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
3036 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
3037 // CHECK3-NEXT:  entry:
3038 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3039 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3040 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3041 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
3042 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
3043 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3044 // CHECK3-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
3045 // CHECK3-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
3046 // CHECK3-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
3047 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3048 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3049 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3050 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
3051 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
3052 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3053 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3054 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3055 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
3056 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3057 // CHECK3-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
3058 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
3059 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
3060 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
3061 // CHECK3-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
3062 // CHECK3-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
3063 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
3064 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
3065 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
3066 // CHECK3-NEXT:    [[A_CASTED17:%.*]] = alloca i32, align 4
3067 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [9 x i8*], align 4
3068 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [9 x i8*], align 4
3069 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [9 x i8*], align 4
3070 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
3071 // CHECK3-NEXT:    [[NN:%.*]] = alloca i32, align 4
3072 // CHECK3-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
3073 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
3074 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
3075 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
3076 // CHECK3-NEXT:    [[NN_CASTED32:%.*]] = alloca i32, align 4
3077 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 4
3078 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 4
3079 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 4
3080 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
3081 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3082 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3083 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
3084 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3085 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3086 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
3087 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
3088 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
3089 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
3090 // CHECK3-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
3091 // CHECK3-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
3092 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
3093 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3094 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
3095 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
3096 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3097 // CHECK3-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
3098 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3099 // CHECK3-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
3100 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3101 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3102 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3103 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3104 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3105 // CHECK3-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
3106 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
3107 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3108 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
3109 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
3110 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3111 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
3112 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
3113 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3114 // CHECK3-NEXT:    store i8* null, i8** [[TMP17]], align 4
3115 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3116 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
3117 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
3118 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3119 // CHECK3-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
3120 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
3121 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3122 // CHECK3-NEXT:    store i8* null, i8** [[TMP22]], align 4
3123 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3124 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
3125 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
3126 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3127 // CHECK3-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
3128 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
3129 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3130 // CHECK3-NEXT:    store i8* null, i8** [[TMP27]], align 4
3131 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3132 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3133 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
3134 // CHECK3-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
3135 // CHECK3-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
3136 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
3137 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3138 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
3139 // CHECK3-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
3140 // CHECK3-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3141 // CHECK3-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
3142 // CHECK3-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
3143 // CHECK3-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
3144 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
3145 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
3146 // CHECK3-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
3147 // CHECK3-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
3148 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
3149 // CHECK3-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
3150 // CHECK3-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
3151 // CHECK3-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
3152 // CHECK3-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
3153 // 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)
3154 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
3155 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
3156 // CHECK3-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
3157 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
3158 // CHECK3-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
3159 // CHECK3-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
3160 // CHECK3-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
3161 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
3162 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
3163 // CHECK3-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
3164 // CHECK3-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
3165 // CHECK3-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP36]])
3166 // CHECK3-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
3167 // CHECK3-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
3168 // CHECK3-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
3169 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
3170 // CHECK3-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
3171 // CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
3172 // CHECK3-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
3173 // CHECK3-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
3174 // CHECK3-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
3175 // CHECK3-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
3176 // CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
3177 // CHECK3-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
3178 // CHECK3-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
3179 // CHECK3-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
3180 // CHECK3-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
3181 // CHECK3-NEXT:    store i8* null, i8** [[TMP63]], align 4
3182 // CHECK3-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
3183 // CHECK3-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
3184 // CHECK3-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
3185 // CHECK3-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
3186 // CHECK3-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3187 // CHECK3:       omp_offload.failed:
3188 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i32 [[TMP58]]) #[[ATTR3]]
3189 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3190 // CHECK3:       omp_offload.cont:
3191 // CHECK3-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
3192 // CHECK3-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
3193 // CHECK3-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
3194 // CHECK3-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
3195 // CHECK3-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
3196 // CHECK3-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
3197 // CHECK3-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
3198 // CHECK3-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
3199 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
3200 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3201 // CHECK3:       omp_if.then:
3202 // CHECK3-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
3203 // CHECK3-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
3204 // CHECK3-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
3205 // CHECK3-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
3206 // CHECK3-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
3207 // CHECK3-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
3208 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
3209 // CHECK3-NEXT:    store i8* null, i8** [[TMP77]], align 4
3210 // CHECK3-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
3211 // CHECK3-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
3212 // CHECK3-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
3213 // CHECK3-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
3214 // CHECK3-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
3215 // CHECK3-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
3216 // CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
3217 // CHECK3-NEXT:    store i8* null, i8** [[TMP82]], align 4
3218 // CHECK3-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
3219 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
3220 // CHECK3-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
3221 // CHECK3-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
3222 // CHECK3-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
3223 // CHECK3:       omp_offload.failed15:
3224 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
3225 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
3226 // CHECK3:       omp_offload.cont16:
3227 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3228 // CHECK3:       omp_if.else:
3229 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
3230 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3231 // CHECK3:       omp_if.end:
3232 // CHECK3-NEXT:    [[TMP87:%.*]] = load i32, i32* [[A]], align 4
3233 // CHECK3-NEXT:    store i32 [[TMP87]], i32* [[A_CASTED17]], align 4
3234 // CHECK3-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A_CASTED17]], align 4
3235 // CHECK3-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
3236 // CHECK3-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP89]], 20
3237 // CHECK3-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
3238 // CHECK3:       omp_if.then19:
3239 // CHECK3-NEXT:    [[TMP90:%.*]] = mul nuw i32 [[TMP1]], 4
3240 // CHECK3-NEXT:    [[TMP91:%.*]] = sext i32 [[TMP90]] to i64
3241 // CHECK3-NEXT:    [[TMP92:%.*]] = mul nuw i32 5, [[TMP3]]
3242 // CHECK3-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP92]], 8
3243 // CHECK3-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
3244 // CHECK3-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3245 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP95]], i8* align 4 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i32 72, i1 false)
3246 // CHECK3-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
3247 // CHECK3-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i32*
3248 // CHECK3-NEXT:    store i32 [[TMP88]], i32* [[TMP97]], align 4
3249 // CHECK3-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
3250 // CHECK3-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
3251 // CHECK3-NEXT:    store i32 [[TMP88]], i32* [[TMP99]], align 4
3252 // CHECK3-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
3253 // CHECK3-NEXT:    store i8* null, i8** [[TMP100]], align 4
3254 // CHECK3-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
3255 // CHECK3-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
3256 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 4
3257 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
3258 // CHECK3-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
3259 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 4
3260 // CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
3261 // CHECK3-NEXT:    store i8* null, i8** [[TMP105]], align 4
3262 // CHECK3-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
3263 // CHECK3-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i32*
3264 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP107]], align 4
3265 // CHECK3-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
3266 // CHECK3-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i32*
3267 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP109]], align 4
3268 // CHECK3-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 2
3269 // CHECK3-NEXT:    store i8* null, i8** [[TMP110]], align 4
3270 // CHECK3-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
3271 // CHECK3-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
3272 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP112]], align 4
3273 // CHECK3-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
3274 // CHECK3-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
3275 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP114]], align 4
3276 // CHECK3-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
3277 // CHECK3-NEXT:    store i64 [[TMP91]], i64* [[TMP115]], align 4
3278 // CHECK3-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 3
3279 // CHECK3-NEXT:    store i8* null, i8** [[TMP116]], align 4
3280 // CHECK3-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
3281 // CHECK3-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
3282 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 4
3283 // CHECK3-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
3284 // CHECK3-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
3285 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 4
3286 // CHECK3-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 4
3287 // CHECK3-NEXT:    store i8* null, i8** [[TMP121]], align 4
3288 // CHECK3-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
3289 // CHECK3-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i32*
3290 // CHECK3-NEXT:    store i32 5, i32* [[TMP123]], align 4
3291 // CHECK3-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
3292 // CHECK3-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i32*
3293 // CHECK3-NEXT:    store i32 5, i32* [[TMP125]], align 4
3294 // CHECK3-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 5
3295 // CHECK3-NEXT:    store i8* null, i8** [[TMP126]], align 4
3296 // CHECK3-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
3297 // CHECK3-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
3298 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP128]], align 4
3299 // CHECK3-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
3300 // CHECK3-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i32*
3301 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP130]], align 4
3302 // CHECK3-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 6
3303 // CHECK3-NEXT:    store i8* null, i8** [[TMP131]], align 4
3304 // CHECK3-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
3305 // CHECK3-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
3306 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 4
3307 // CHECK3-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
3308 // CHECK3-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
3309 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 4
3310 // CHECK3-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
3311 // CHECK3-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 4
3312 // CHECK3-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 7
3313 // CHECK3-NEXT:    store i8* null, i8** [[TMP137]], align 4
3314 // CHECK3-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
3315 // CHECK3-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
3316 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 4
3317 // CHECK3-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
3318 // CHECK3-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
3319 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 4
3320 // CHECK3-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 8
3321 // CHECK3-NEXT:    store i8* null, i8** [[TMP142]], align 4
3322 // CHECK3-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
3323 // CHECK3-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
3324 // CHECK3-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3325 // CHECK3-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3326 // CHECK3-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
3327 // CHECK3-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
3328 // CHECK3:       omp_offload.failed23:
3329 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
3330 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
3331 // CHECK3:       omp_offload.cont24:
3332 // CHECK3-NEXT:    br label [[OMP_IF_END26:%.*]]
3333 // CHECK3:       omp_if.else25:
3334 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
3335 // CHECK3-NEXT:    br label [[OMP_IF_END26]]
3336 // CHECK3:       omp_if.end26:
3337 // CHECK3-NEXT:    store i32 0, i32* [[NN]], align 4
3338 // CHECK3-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
3339 // CHECK3-NEXT:    store i32 [[TMP148]], i32* [[NN_CASTED]], align 4
3340 // CHECK3-NEXT:    [[TMP149:%.*]] = load i32, i32* [[NN_CASTED]], align 4
3341 // CHECK3-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
3342 // CHECK3-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i32*
3343 // CHECK3-NEXT:    store i32 [[TMP149]], i32* [[TMP151]], align 4
3344 // CHECK3-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
3345 // CHECK3-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
3346 // CHECK3-NEXT:    store i32 [[TMP149]], i32* [[TMP153]], align 4
3347 // CHECK3-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
3348 // CHECK3-NEXT:    store i8* null, i8** [[TMP154]], align 4
3349 // CHECK3-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
3350 // CHECK3-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
3351 // CHECK3-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3352 // CHECK3-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
3353 // CHECK3-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED30:%.*]], label [[OMP_OFFLOAD_CONT31:%.*]]
3354 // CHECK3:       omp_offload.failed30:
3355 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i32 [[TMP149]]) #[[ATTR3]]
3356 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT31]]
3357 // CHECK3:       omp_offload.cont31:
3358 // CHECK3-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
3359 // CHECK3-NEXT:    store i32 [[TMP159]], i32* [[NN_CASTED32]], align 4
3360 // CHECK3-NEXT:    [[TMP160:%.*]] = load i32, i32* [[NN_CASTED32]], align 4
3361 // CHECK3-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
3362 // CHECK3-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i32*
3363 // CHECK3-NEXT:    store i32 [[TMP160]], i32* [[TMP162]], align 4
3364 // CHECK3-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
3365 // CHECK3-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i32*
3366 // CHECK3-NEXT:    store i32 [[TMP160]], i32* [[TMP164]], align 4
3367 // CHECK3-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i32 0, i32 0
3368 // CHECK3-NEXT:    store i8* null, i8** [[TMP165]], align 4
3369 // CHECK3-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
3370 // CHECK3-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
3371 // CHECK3-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3372 // CHECK3-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
3373 // CHECK3-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
3374 // CHECK3:       omp_offload.failed36:
3375 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i32 [[TMP160]]) #[[ATTR3]]
3376 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
3377 // CHECK3:       omp_offload.cont37:
3378 // CHECK3-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
3379 // CHECK3-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3380 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
3381 // CHECK3-NEXT:    ret i32 [[TMP170]]
3382 //
3383 //
3384 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
3385 // CHECK3-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
3386 // CHECK3-NEXT:  entry:
3387 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3388 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
3389 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
3390 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3391 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
3392 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3393 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3394 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
3395 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3396 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3397 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
3398 // CHECK3-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
3399 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
3400 // CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3401 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
3402 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3403 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
3404 // CHECK3-NEXT:    ret void
3405 //
3406 //
3407 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
3408 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
3409 // CHECK3-NEXT:  entry:
3410 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3411 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3412 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3413 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3414 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3415 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3416 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3417 // CHECK3-NEXT:    ret void
3418 //
3419 //
3420 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_privates_map.
3421 // 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]+]] {
3422 // CHECK3-NEXT:  entry:
3423 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
3424 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
3425 // CHECK3-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
3426 // CHECK3-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
3427 // CHECK3-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
3428 // CHECK3-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
3429 // CHECK3-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
3430 // CHECK3-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
3431 // CHECK3-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
3432 // CHECK3-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
3433 // CHECK3-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
3434 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
3435 // CHECK3-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
3436 // CHECK3-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
3437 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
3438 // CHECK3-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
3439 // CHECK3-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
3440 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
3441 // CHECK3-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
3442 // CHECK3-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
3443 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
3444 // CHECK3-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
3445 // CHECK3-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
3446 // CHECK3-NEXT:    ret void
3447 //
3448 //
3449 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
3450 // CHECK3-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
3451 // CHECK3-NEXT:  entry:
3452 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
3453 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
3454 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
3455 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
3456 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
3457 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
3458 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
3459 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
3460 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
3461 // CHECK3-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
3462 // CHECK3-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
3463 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
3464 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
3465 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
3466 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
3467 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
3468 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
3469 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
3470 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
3471 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
3472 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
3473 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
3474 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
3475 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
3476 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
3477 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
3478 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
3479 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
3480 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
3481 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
3482 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
3483 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
3484 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !25
3485 // CHECK3-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
3486 // 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 !25
3487 // CHECK3-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !25
3488 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
3489 // CHECK3-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
3490 // CHECK3-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !25
3491 // CHECK3-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
3492 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
3493 // 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]]
3494 // CHECK3-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !25
3495 // CHECK3-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !25
3496 // CHECK3-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !25
3497 // CHECK3-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !25
3498 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
3499 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
3500 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
3501 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
3502 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
3503 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
3504 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
3505 // CHECK3-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
3506 // CHECK3-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
3507 // CHECK3-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
3508 // CHECK3:       omp_offload.failed.i:
3509 // CHECK3-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
3510 // CHECK3-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
3511 // CHECK3-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !25
3512 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !25
3513 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
3514 // CHECK3-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
3515 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
3516 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
3517 // CHECK3-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
3518 // CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
3519 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
3520 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
3521 // CHECK3:       .omp_outlined..1.exit:
3522 // CHECK3-NEXT:    ret i32 0
3523 //
3524 //
3525 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
3526 // CHECK3-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
3527 // CHECK3-NEXT:  entry:
3528 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3529 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3530 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3531 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
3532 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
3533 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
3534 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
3535 // CHECK3-NEXT:    ret void
3536 //
3537 //
3538 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
3539 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
3540 // CHECK3-NEXT:  entry:
3541 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3542 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3543 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3544 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3545 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3546 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3547 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
3548 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
3549 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
3550 // CHECK3-NEXT:    ret void
3551 //
3552 //
3553 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
3554 // CHECK3-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
3555 // CHECK3-NEXT:  entry:
3556 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3557 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3558 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3559 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3560 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
3561 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3562 // CHECK3-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
3563 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3564 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
3565 // CHECK3-NEXT:    ret void
3566 //
3567 //
3568 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
3569 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
3570 // CHECK3-NEXT:  entry:
3571 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3572 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3573 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3574 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3575 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3576 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3577 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3578 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
3579 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
3580 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
3581 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
3582 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
3583 // CHECK3-NEXT:    ret void
3584 //
3585 //
3586 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
3587 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
3588 // CHECK3-NEXT:  entry:
3589 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3590 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3591 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3592 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3593 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3594 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3595 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3596 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
3597 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
3598 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
3599 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
3600 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3601 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
3602 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3603 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
3604 // CHECK3-NEXT:    ret void
3605 //
3606 //
3607 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
3608 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
3609 // CHECK3-NEXT:  entry:
3610 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3611 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3612 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3613 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3614 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3615 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3616 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3617 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3618 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3619 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
3620 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
3621 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
3622 // CHECK3-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
3623 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
3624 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
3625 // CHECK3-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
3626 // CHECK3-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
3627 // CHECK3-NEXT:    ret void
3628 //
3629 //
3630 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
3631 // 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:%.*]]) #[[ATTR2]] {
3632 // CHECK3-NEXT:  entry:
3633 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3634 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
3635 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3636 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
3637 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
3638 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3639 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
3640 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
3641 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
3642 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3643 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3644 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
3645 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3646 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
3647 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
3648 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3649 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
3650 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
3651 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
3652 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
3653 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3654 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
3655 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
3656 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3657 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
3658 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
3659 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
3660 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
3661 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
3662 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
3663 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
3664 // CHECK3-NEXT:    ret void
3665 //
3666 //
3667 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
3668 // 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:%.*]]) #[[ATTR2]] {
3669 // CHECK3-NEXT:  entry:
3670 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3671 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3672 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3673 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
3674 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3675 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
3676 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
3677 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3678 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
3679 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
3680 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
3681 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3682 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3683 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3684 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
3685 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3686 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
3687 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
3688 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3689 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
3690 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
3691 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
3692 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
3693 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3694 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
3695 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
3696 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3697 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
3698 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
3699 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
3700 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
3701 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
3702 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
3703 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
3704 // CHECK3-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
3705 // CHECK3-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
3706 // CHECK3-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
3707 // CHECK3-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
3708 // CHECK3-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
3709 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
3710 // CHECK3-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
3711 // CHECK3-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
3712 // CHECK3-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
3713 // CHECK3-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
3714 // CHECK3-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
3715 // CHECK3-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
3716 // CHECK3-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
3717 // CHECK3-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
3718 // CHECK3-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
3719 // CHECK3-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
3720 // CHECK3-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
3721 // CHECK3-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
3722 // CHECK3-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
3723 // CHECK3-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
3724 // CHECK3-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
3725 // CHECK3-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
3726 // CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
3727 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
3728 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
3729 // CHECK3-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
3730 // CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
3731 // CHECK3-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
3732 // CHECK3-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
3733 // CHECK3-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
3734 // CHECK3-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
3735 // CHECK3-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
3736 // CHECK3-NEXT:    ret void
3737 //
3738 //
3739 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
3740 // CHECK3-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
3741 // CHECK3-NEXT:  entry:
3742 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
3743 // CHECK3-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
3744 // CHECK3-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
3745 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
3746 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
3747 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
3748 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP1]])
3749 // CHECK3-NEXT:    ret void
3750 //
3751 //
3752 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
3753 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
3754 // CHECK3-NEXT:  entry:
3755 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3756 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3757 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
3758 // CHECK3-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
3759 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3760 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3761 // CHECK3-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
3762 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
3763 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
3764 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
3765 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP1]])
3766 // CHECK3-NEXT:    ret void
3767 //
3768 //
3769 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..13
3770 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
3771 // CHECK3-NEXT:  entry:
3772 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3773 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3774 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
3775 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3776 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3777 // CHECK3-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
3778 // CHECK3-NEXT:    ret void
3779 //
3780 //
3781 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
3782 // CHECK3-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
3783 // CHECK3-NEXT:  entry:
3784 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
3785 // CHECK3-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
3786 // CHECK3-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
3787 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
3788 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
3789 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
3790 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP1]])
3791 // CHECK3-NEXT:    ret void
3792 //
3793 //
3794 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..16
3795 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
3796 // CHECK3-NEXT:  entry:
3797 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3798 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3799 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
3800 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3801 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3802 // CHECK3-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
3803 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
3804 // CHECK3-NEXT:    ret void
3805 //
3806 //
3807 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..17
3808 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
3809 // CHECK3-NEXT:  entry:
3810 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3811 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3812 // CHECK3-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
3813 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3814 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3815 // CHECK3-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
3816 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
3817 // CHECK3-NEXT:    ret void
3818 //
3819 //
3820 // CHECK3-LABEL: define {{[^@]+}}@_Z6bazzzziPi
3821 // CHECK3-SAME: (i32 noundef [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
3822 // CHECK3-NEXT:  entry:
3823 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3824 // CHECK3-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 4
3825 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
3826 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
3827 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
3828 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3829 // CHECK3-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 4
3830 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3831 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3832 // CHECK3-NEXT:    [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32*
3833 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[TMP2]], align 4
3834 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3835 // CHECK3-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
3836 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[TMP4]], align 4
3837 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3838 // CHECK3-NEXT:    store i8* null, i8** [[TMP5]], align 4
3839 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3840 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3841 // CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3842 // CHECK3-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
3843 // CHECK3-NEXT:    br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3844 // CHECK3:       omp_offload.failed:
3845 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i32 [[TMP0]]) #[[ATTR3]]
3846 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3847 // CHECK3:       omp_offload.cont:
3848 // CHECK3-NEXT:    ret void
3849 //
3850 //
3851 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
3852 // CHECK3-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
3853 // CHECK3-NEXT:  entry:
3854 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3855 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3856 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3857 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i32 [[TMP0]])
3858 // CHECK3-NEXT:    ret void
3859 //
3860 //
3861 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..20
3862 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
3863 // CHECK3-NEXT:  entry:
3864 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3865 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3866 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3867 // CHECK3-NEXT:    [[F:%.*]] = alloca i32*, align 4
3868 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3869 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3870 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3871 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3872 // CHECK3-NEXT:    ret void
3873 //
3874 //
3875 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
3876 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
3877 // CHECK3-NEXT:  entry:
3878 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3879 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3880 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
3881 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3882 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3883 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3884 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
3885 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
3886 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
3887 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
3888 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
3889 // CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
3890 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
3891 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
3892 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
3893 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3894 // CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
3895 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
3896 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
3897 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
3898 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
3899 // CHECK3-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
3900 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
3901 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
3902 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
3903 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
3904 // CHECK3-NEXT:    ret i32 [[TMP8]]
3905 //
3906 //
3907 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
3908 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
3909 // CHECK3-NEXT:  entry:
3910 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3911 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3912 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
3913 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
3914 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3915 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
3916 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
3917 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
3918 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
3919 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
3920 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3921 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3922 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3923 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3924 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
3925 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
3926 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3927 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3928 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
3929 // CHECK3-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
3930 // CHECK3-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
3931 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
3932 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
3933 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
3934 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
3935 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
3936 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
3937 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3938 // CHECK3:       omp_if.then:
3939 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
3940 // CHECK3-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
3941 // CHECK3-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
3942 // CHECK3-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
3943 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3944 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i32 40, i1 false)
3945 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3946 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
3947 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
3948 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3949 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
3950 // CHECK3-NEXT:    store double* [[A]], double** [[TMP14]], align 4
3951 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3952 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
3953 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3954 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
3955 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
3956 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3957 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
3958 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
3959 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3960 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
3961 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3962 // CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
3963 // CHECK3-NEXT:    store i32 2, i32* [[TMP22]], align 4
3964 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3965 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
3966 // CHECK3-NEXT:    store i32 2, i32* [[TMP24]], align 4
3967 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3968 // CHECK3-NEXT:    store i8* null, i8** [[TMP25]], align 4
3969 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
3970 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
3971 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
3972 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
3973 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
3974 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
3975 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
3976 // CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
3977 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
3978 // CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
3979 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
3980 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
3981 // CHECK3-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
3982 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
3983 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
3984 // CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
3985 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
3986 // CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
3987 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3988 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3989 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3990 // CHECK3-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3991 // CHECK3-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
3992 // CHECK3-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3993 // CHECK3:       omp_offload.failed:
3994 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
3995 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3996 // CHECK3:       omp_offload.cont:
3997 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3998 // CHECK3:       omp_if.else:
3999 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
4000 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4001 // CHECK3:       omp_if.end:
4002 // CHECK3-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
4003 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
4004 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
4005 // CHECK3-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
4006 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
4007 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
4008 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
4009 // CHECK3-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4010 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
4011 // CHECK3-NEXT:    ret i32 [[ADD3]]
4012 //
4013 //
4014 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
4015 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
4016 // CHECK3-NEXT:  entry:
4017 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4018 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
4019 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
4020 // CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4021 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4022 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4023 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4024 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
4025 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
4026 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
4027 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
4028 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4029 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
4030 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
4031 // CHECK3-NEXT:    store i8 0, i8* [[AAA]], align 1
4032 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4033 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4034 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4035 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4036 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4037 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
4038 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4039 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
4040 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
4041 // CHECK3-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
4042 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
4043 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4044 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
4045 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4046 // CHECK3:       omp_if.then:
4047 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4048 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
4049 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
4050 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4051 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4052 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
4053 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4054 // CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
4055 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4056 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4057 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
4058 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4059 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4060 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
4061 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4062 // CHECK3-NEXT:    store i8* null, i8** [[TMP16]], align 4
4063 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4064 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
4065 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
4066 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4067 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4068 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
4069 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4070 // CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
4071 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4072 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
4073 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
4074 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4075 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
4076 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
4077 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4078 // CHECK3-NEXT:    store i8* null, i8** [[TMP26]], align 4
4079 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4080 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4081 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4082 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
4083 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4084 // CHECK3:       omp_offload.failed:
4085 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
4086 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4087 // CHECK3:       omp_offload.cont:
4088 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4089 // CHECK3:       omp_if.else:
4090 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
4091 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4092 // CHECK3:       omp_if.end:
4093 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
4094 // CHECK3-NEXT:    ret i32 [[TMP31]]
4095 //
4096 //
4097 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4098 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
4099 // CHECK3-NEXT:  entry:
4100 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4101 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
4102 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
4103 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4104 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4105 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4106 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4107 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4108 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4109 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4110 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
4111 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
4112 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4113 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4114 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4115 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4116 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4117 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
4118 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4119 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4120 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
4121 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4122 // CHECK3:       omp_if.then:
4123 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4124 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
4125 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
4126 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4127 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
4128 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
4129 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4130 // CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
4131 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4132 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
4133 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
4134 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4135 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4136 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
4137 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4138 // CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
4139 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4140 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
4141 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
4142 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4143 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
4144 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
4145 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4146 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
4147 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4148 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4149 // CHECK3-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4150 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
4151 // CHECK3-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4152 // CHECK3:       omp_offload.failed:
4153 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4154 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4155 // CHECK3:       omp_offload.cont:
4156 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4157 // CHECK3:       omp_if.else:
4158 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
4159 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4160 // CHECK3:       omp_if.end:
4161 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
4162 // CHECK3-NEXT:    ret i32 [[TMP24]]
4163 //
4164 //
4165 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
4166 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
4167 // CHECK3-NEXT:  entry:
4168 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4169 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4170 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4171 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4172 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
4173 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
4174 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4175 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4176 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4177 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4178 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
4179 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4180 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4181 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4182 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
4183 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
4184 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
4185 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
4186 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
4187 // CHECK3-NEXT:    ret void
4188 //
4189 //
4190 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..23
4191 // 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]] {
4192 // CHECK3-NEXT:  entry:
4193 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4194 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4195 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
4196 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
4197 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4198 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4199 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
4200 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4201 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4202 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
4203 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
4204 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4205 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4206 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
4207 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
4208 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4209 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4210 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
4211 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
4212 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
4213 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
4214 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
4215 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4
4216 // CHECK3-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
4217 // CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
4218 // CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
4219 // CHECK3-NEXT:    store double [[INC]], double* [[A3]], align 4
4220 // CHECK3-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
4221 // CHECK3-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
4222 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
4223 // CHECK3-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
4224 // CHECK3-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
4225 // CHECK3-NEXT:    ret void
4226 //
4227 //
4228 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
4229 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4230 // CHECK3-NEXT:  entry:
4231 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4232 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4233 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
4234 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4235 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4236 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4237 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
4238 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4239 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4240 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
4241 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4242 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4243 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
4244 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4245 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
4246 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
4247 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
4248 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
4249 // CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4250 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
4251 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4252 // CHECK3-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
4253 // CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
4254 // CHECK3-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
4255 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
4256 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
4257 // CHECK3-NEXT:    ret void
4258 //
4259 //
4260 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..26
4261 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4262 // CHECK3-NEXT:  entry:
4263 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4264 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4265 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4266 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4267 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
4268 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4269 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4270 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4271 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4272 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4273 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
4274 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4275 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4276 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
4277 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4278 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
4279 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
4280 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
4281 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
4282 // CHECK3-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
4283 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
4284 // CHECK3-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
4285 // CHECK3-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
4286 // CHECK3-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
4287 // CHECK3-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
4288 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
4289 // CHECK3-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
4290 // CHECK3-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
4291 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
4292 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4293 // CHECK3-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
4294 // CHECK3-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
4295 // CHECK3-NEXT:    ret void
4296 //
4297 //
4298 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
4299 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4300 // CHECK3-NEXT:  entry:
4301 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4302 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4303 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4304 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4305 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4306 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4307 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4308 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4309 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4310 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4311 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
4312 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
4313 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
4314 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
4315 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4316 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
4317 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4318 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
4319 // CHECK3-NEXT:    ret void
4320 //
4321 //
4322 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..29
4323 // 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]] {
4324 // CHECK3-NEXT:  entry:
4325 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4326 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4327 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4328 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4329 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
4330 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4331 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4332 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4333 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4334 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
4335 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4336 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
4337 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
4338 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
4339 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
4340 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
4341 // CHECK3-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
4342 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
4343 // CHECK3-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
4344 // CHECK3-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
4345 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
4346 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4347 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
4348 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
4349 // CHECK3-NEXT:    ret void
4350 //
4351 //
4352 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4353 // CHECK3-SAME: () #[[ATTR4]] {
4354 // CHECK3-NEXT:  entry:
4355 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
4356 // CHECK3-NEXT:    ret void
4357 //
4358 //
4359 // CHECK4-LABEL: define {{[^@]+}}@_Z3fooi
4360 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
4361 // CHECK4-NEXT:  entry:
4362 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4363 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
4364 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
4365 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
4366 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
4367 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
4368 // CHECK4-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
4369 // CHECK4-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
4370 // CHECK4-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
4371 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4372 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4373 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4374 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4375 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
4376 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4377 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4378 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4379 // CHECK4-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
4380 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4381 // CHECK4-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
4382 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
4383 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
4384 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
4385 // CHECK4-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
4386 // CHECK4-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
4387 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
4388 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
4389 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
4390 // CHECK4-NEXT:    [[A_CASTED17:%.*]] = alloca i32, align 4
4391 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [9 x i8*], align 4
4392 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [9 x i8*], align 4
4393 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [9 x i8*], align 4
4394 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
4395 // CHECK4-NEXT:    [[NN:%.*]] = alloca i32, align 4
4396 // CHECK4-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
4397 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
4398 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
4399 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
4400 // CHECK4-NEXT:    [[NN_CASTED32:%.*]] = alloca i32, align 4
4401 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 4
4402 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 4
4403 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 4
4404 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
4405 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4406 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
4407 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
4408 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
4409 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
4410 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
4411 // CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
4412 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
4413 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
4414 // CHECK4-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
4415 // CHECK4-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
4416 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
4417 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4418 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
4419 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
4420 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4421 // CHECK4-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
4422 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4423 // CHECK4-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
4424 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4425 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4426 // CHECK4-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4427 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4428 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4429 // CHECK4-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
4430 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
4431 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4432 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
4433 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
4434 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4435 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
4436 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
4437 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4438 // CHECK4-NEXT:    store i8* null, i8** [[TMP17]], align 4
4439 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4440 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
4441 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
4442 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4443 // CHECK4-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
4444 // CHECK4-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
4445 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4446 // CHECK4-NEXT:    store i8* null, i8** [[TMP22]], align 4
4447 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4448 // CHECK4-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
4449 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
4450 // CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4451 // CHECK4-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
4452 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
4453 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4454 // CHECK4-NEXT:    store i8* null, i8** [[TMP27]], align 4
4455 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4456 // CHECK4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4457 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
4458 // CHECK4-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
4459 // CHECK4-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
4460 // CHECK4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
4461 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4462 // CHECK4-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
4463 // CHECK4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
4464 // CHECK4-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4465 // CHECK4-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
4466 // CHECK4-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
4467 // CHECK4-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
4468 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
4469 // CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
4470 // CHECK4-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
4471 // CHECK4-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
4472 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
4473 // CHECK4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
4474 // CHECK4-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
4475 // CHECK4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
4476 // CHECK4-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
4477 // 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)
4478 // CHECK4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
4479 // CHECK4-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
4480 // CHECK4-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
4481 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
4482 // CHECK4-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
4483 // CHECK4-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
4484 // CHECK4-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
4485 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
4486 // CHECK4-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
4487 // CHECK4-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
4488 // CHECK4-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
4489 // CHECK4-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP36]])
4490 // CHECK4-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
4491 // CHECK4-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
4492 // CHECK4-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
4493 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
4494 // CHECK4-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
4495 // CHECK4-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
4496 // CHECK4-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
4497 // CHECK4-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
4498 // CHECK4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
4499 // CHECK4-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
4500 // CHECK4-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
4501 // CHECK4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
4502 // CHECK4-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
4503 // CHECK4-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
4504 // CHECK4-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
4505 // CHECK4-NEXT:    store i8* null, i8** [[TMP63]], align 4
4506 // CHECK4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
4507 // CHECK4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
4508 // CHECK4-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
4509 // CHECK4-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
4510 // CHECK4-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4511 // CHECK4:       omp_offload.failed:
4512 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i32 [[TMP58]]) #[[ATTR3]]
4513 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4514 // CHECK4:       omp_offload.cont:
4515 // CHECK4-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
4516 // CHECK4-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
4517 // CHECK4-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
4518 // CHECK4-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
4519 // CHECK4-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
4520 // CHECK4-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
4521 // CHECK4-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
4522 // CHECK4-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
4523 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
4524 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4525 // CHECK4:       omp_if.then:
4526 // CHECK4-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
4527 // CHECK4-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
4528 // CHECK4-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
4529 // CHECK4-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
4530 // CHECK4-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
4531 // CHECK4-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
4532 // CHECK4-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
4533 // CHECK4-NEXT:    store i8* null, i8** [[TMP77]], align 4
4534 // CHECK4-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
4535 // CHECK4-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
4536 // CHECK4-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
4537 // CHECK4-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
4538 // CHECK4-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
4539 // CHECK4-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
4540 // CHECK4-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
4541 // CHECK4-NEXT:    store i8* null, i8** [[TMP82]], align 4
4542 // CHECK4-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
4543 // CHECK4-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
4544 // CHECK4-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
4545 // CHECK4-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
4546 // CHECK4-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
4547 // CHECK4:       omp_offload.failed15:
4548 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
4549 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
4550 // CHECK4:       omp_offload.cont16:
4551 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
4552 // CHECK4:       omp_if.else:
4553 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
4554 // CHECK4-NEXT:    br label [[OMP_IF_END]]
4555 // CHECK4:       omp_if.end:
4556 // CHECK4-NEXT:    [[TMP87:%.*]] = load i32, i32* [[A]], align 4
4557 // CHECK4-NEXT:    store i32 [[TMP87]], i32* [[A_CASTED17]], align 4
4558 // CHECK4-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A_CASTED17]], align 4
4559 // CHECK4-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
4560 // CHECK4-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP89]], 20
4561 // CHECK4-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
4562 // CHECK4:       omp_if.then19:
4563 // CHECK4-NEXT:    [[TMP90:%.*]] = mul nuw i32 [[TMP1]], 4
4564 // CHECK4-NEXT:    [[TMP91:%.*]] = sext i32 [[TMP90]] to i64
4565 // CHECK4-NEXT:    [[TMP92:%.*]] = mul nuw i32 5, [[TMP3]]
4566 // CHECK4-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP92]], 8
4567 // CHECK4-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
4568 // CHECK4-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
4569 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP95]], i8* align 4 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i32 72, i1 false)
4570 // CHECK4-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
4571 // CHECK4-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i32*
4572 // CHECK4-NEXT:    store i32 [[TMP88]], i32* [[TMP97]], align 4
4573 // CHECK4-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
4574 // CHECK4-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
4575 // CHECK4-NEXT:    store i32 [[TMP88]], i32* [[TMP99]], align 4
4576 // CHECK4-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
4577 // CHECK4-NEXT:    store i8* null, i8** [[TMP100]], align 4
4578 // CHECK4-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
4579 // CHECK4-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
4580 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 4
4581 // CHECK4-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
4582 // CHECK4-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
4583 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 4
4584 // CHECK4-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
4585 // CHECK4-NEXT:    store i8* null, i8** [[TMP105]], align 4
4586 // CHECK4-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
4587 // CHECK4-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i32*
4588 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP107]], align 4
4589 // CHECK4-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
4590 // CHECK4-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i32*
4591 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP109]], align 4
4592 // CHECK4-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 2
4593 // CHECK4-NEXT:    store i8* null, i8** [[TMP110]], align 4
4594 // CHECK4-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
4595 // CHECK4-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
4596 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP112]], align 4
4597 // CHECK4-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
4598 // CHECK4-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
4599 // CHECK4-NEXT:    store float* [[VLA]], float** [[TMP114]], align 4
4600 // CHECK4-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
4601 // CHECK4-NEXT:    store i64 [[TMP91]], i64* [[TMP115]], align 4
4602 // CHECK4-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 3
4603 // CHECK4-NEXT:    store i8* null, i8** [[TMP116]], align 4
4604 // CHECK4-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
4605 // CHECK4-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
4606 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 4
4607 // CHECK4-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
4608 // CHECK4-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
4609 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 4
4610 // CHECK4-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 4
4611 // CHECK4-NEXT:    store i8* null, i8** [[TMP121]], align 4
4612 // CHECK4-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
4613 // CHECK4-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i32*
4614 // CHECK4-NEXT:    store i32 5, i32* [[TMP123]], align 4
4615 // CHECK4-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
4616 // CHECK4-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i32*
4617 // CHECK4-NEXT:    store i32 5, i32* [[TMP125]], align 4
4618 // CHECK4-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 5
4619 // CHECK4-NEXT:    store i8* null, i8** [[TMP126]], align 4
4620 // CHECK4-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
4621 // CHECK4-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
4622 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP128]], align 4
4623 // CHECK4-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
4624 // CHECK4-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i32*
4625 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP130]], align 4
4626 // CHECK4-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 6
4627 // CHECK4-NEXT:    store i8* null, i8** [[TMP131]], align 4
4628 // CHECK4-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
4629 // CHECK4-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
4630 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 4
4631 // CHECK4-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
4632 // CHECK4-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
4633 // CHECK4-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 4
4634 // CHECK4-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
4635 // CHECK4-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 4
4636 // CHECK4-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 7
4637 // CHECK4-NEXT:    store i8* null, i8** [[TMP137]], align 4
4638 // CHECK4-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
4639 // CHECK4-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
4640 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 4
4641 // CHECK4-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
4642 // CHECK4-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
4643 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 4
4644 // CHECK4-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 8
4645 // CHECK4-NEXT:    store i8* null, i8** [[TMP142]], align 4
4646 // CHECK4-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
4647 // CHECK4-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
4648 // CHECK4-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4649 // CHECK4-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4650 // CHECK4-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
4651 // CHECK4-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
4652 // CHECK4:       omp_offload.failed23:
4653 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
4654 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
4655 // CHECK4:       omp_offload.cont24:
4656 // CHECK4-NEXT:    br label [[OMP_IF_END26:%.*]]
4657 // CHECK4:       omp_if.else25:
4658 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
4659 // CHECK4-NEXT:    br label [[OMP_IF_END26]]
4660 // CHECK4:       omp_if.end26:
4661 // CHECK4-NEXT:    store i32 0, i32* [[NN]], align 4
4662 // CHECK4-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
4663 // CHECK4-NEXT:    store i32 [[TMP148]], i32* [[NN_CASTED]], align 4
4664 // CHECK4-NEXT:    [[TMP149:%.*]] = load i32, i32* [[NN_CASTED]], align 4
4665 // CHECK4-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
4666 // CHECK4-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i32*
4667 // CHECK4-NEXT:    store i32 [[TMP149]], i32* [[TMP151]], align 4
4668 // CHECK4-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
4669 // CHECK4-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
4670 // CHECK4-NEXT:    store i32 [[TMP149]], i32* [[TMP153]], align 4
4671 // CHECK4-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
4672 // CHECK4-NEXT:    store i8* null, i8** [[TMP154]], align 4
4673 // CHECK4-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
4674 // CHECK4-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
4675 // CHECK4-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4676 // CHECK4-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
4677 // CHECK4-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED30:%.*]], label [[OMP_OFFLOAD_CONT31:%.*]]
4678 // CHECK4:       omp_offload.failed30:
4679 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i32 [[TMP149]]) #[[ATTR3]]
4680 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT31]]
4681 // CHECK4:       omp_offload.cont31:
4682 // CHECK4-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
4683 // CHECK4-NEXT:    store i32 [[TMP159]], i32* [[NN_CASTED32]], align 4
4684 // CHECK4-NEXT:    [[TMP160:%.*]] = load i32, i32* [[NN_CASTED32]], align 4
4685 // CHECK4-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
4686 // CHECK4-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i32*
4687 // CHECK4-NEXT:    store i32 [[TMP160]], i32* [[TMP162]], align 4
4688 // CHECK4-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
4689 // CHECK4-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i32*
4690 // CHECK4-NEXT:    store i32 [[TMP160]], i32* [[TMP164]], align 4
4691 // CHECK4-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i32 0, i32 0
4692 // CHECK4-NEXT:    store i8* null, i8** [[TMP165]], align 4
4693 // CHECK4-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
4694 // CHECK4-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
4695 // CHECK4-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4696 // CHECK4-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
4697 // CHECK4-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
4698 // CHECK4:       omp_offload.failed36:
4699 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i32 [[TMP160]]) #[[ATTR3]]
4700 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
4701 // CHECK4:       omp_offload.cont37:
4702 // CHECK4-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
4703 // CHECK4-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4704 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
4705 // CHECK4-NEXT:    ret i32 [[TMP170]]
4706 //
4707 //
4708 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
4709 // CHECK4-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
4710 // CHECK4-NEXT:  entry:
4711 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4712 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4713 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
4714 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4715 // CHECK4-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
4716 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4717 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4718 // CHECK4-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
4719 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4720 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4721 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
4722 // CHECK4-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
4723 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
4724 // CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4725 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
4726 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4727 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
4728 // CHECK4-NEXT:    ret void
4729 //
4730 //
4731 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
4732 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4733 // CHECK4-NEXT:  entry:
4734 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4735 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4736 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4737 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4738 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4739 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4740 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4741 // CHECK4-NEXT:    ret void
4742 //
4743 //
4744 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_privates_map.
4745 // 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]+]] {
4746 // CHECK4-NEXT:  entry:
4747 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
4748 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
4749 // CHECK4-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
4750 // CHECK4-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
4751 // CHECK4-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
4752 // CHECK4-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
4753 // CHECK4-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
4754 // CHECK4-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
4755 // CHECK4-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
4756 // CHECK4-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
4757 // CHECK4-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
4758 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
4759 // CHECK4-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
4760 // CHECK4-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
4761 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
4762 // CHECK4-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
4763 // CHECK4-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
4764 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
4765 // CHECK4-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
4766 // CHECK4-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
4767 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
4768 // CHECK4-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
4769 // CHECK4-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
4770 // CHECK4-NEXT:    ret void
4771 //
4772 //
4773 // CHECK4-LABEL: define {{[^@]+}}@.omp_task_entry.
4774 // CHECK4-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
4775 // CHECK4-NEXT:  entry:
4776 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
4777 // CHECK4-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
4778 // CHECK4-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
4779 // CHECK4-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
4780 // CHECK4-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
4781 // CHECK4-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
4782 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
4783 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
4784 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
4785 // CHECK4-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
4786 // CHECK4-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
4787 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
4788 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
4789 // CHECK4-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
4790 // CHECK4-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
4791 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
4792 // CHECK4-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
4793 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
4794 // CHECK4-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
4795 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
4796 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
4797 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
4798 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
4799 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
4800 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
4801 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
4802 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
4803 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
4804 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
4805 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
4806 // CHECK4-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
4807 // CHECK4-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
4808 // CHECK4-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !25
4809 // CHECK4-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
4810 // 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 !25
4811 // CHECK4-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !25
4812 // CHECK4-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
4813 // CHECK4-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
4814 // CHECK4-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !25
4815 // CHECK4-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
4816 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
4817 // 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]]
4818 // CHECK4-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !25
4819 // CHECK4-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !25
4820 // CHECK4-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !25
4821 // CHECK4-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !25
4822 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
4823 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
4824 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
4825 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
4826 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
4827 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
4828 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
4829 // CHECK4-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
4830 // CHECK4-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4831 // CHECK4-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
4832 // CHECK4:       omp_offload.failed.i:
4833 // CHECK4-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
4834 // CHECK4-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
4835 // CHECK4-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !25
4836 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !25
4837 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
4838 // CHECK4-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
4839 // CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
4840 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
4841 // CHECK4-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
4842 // CHECK4-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
4843 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
4844 // CHECK4-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
4845 // CHECK4:       .omp_outlined..1.exit:
4846 // CHECK4-NEXT:    ret i32 0
4847 //
4848 //
4849 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
4850 // CHECK4-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
4851 // CHECK4-NEXT:  entry:
4852 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4853 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4854 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4855 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4856 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4857 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4858 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
4859 // CHECK4-NEXT:    ret void
4860 //
4861 //
4862 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
4863 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
4864 // CHECK4-NEXT:  entry:
4865 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4866 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4867 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4868 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4869 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4870 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4871 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4872 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
4873 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
4874 // CHECK4-NEXT:    ret void
4875 //
4876 //
4877 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
4878 // CHECK4-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4879 // CHECK4-NEXT:  entry:
4880 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4881 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4882 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4883 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4884 // CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
4885 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4886 // CHECK4-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
4887 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4888 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
4889 // CHECK4-NEXT:    ret void
4890 //
4891 //
4892 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
4893 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4894 // CHECK4-NEXT:  entry:
4895 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4896 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4897 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4898 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4899 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4900 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4901 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4902 // CHECK4-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
4903 // CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
4904 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
4905 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
4906 // CHECK4-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
4907 // CHECK4-NEXT:    ret void
4908 //
4909 //
4910 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
4911 // CHECK4-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4912 // CHECK4-NEXT:  entry:
4913 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4914 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4915 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4916 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
4917 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4918 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4919 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4920 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4921 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
4922 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
4923 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
4924 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
4925 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
4926 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
4927 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
4928 // CHECK4-NEXT:    ret void
4929 //
4930 //
4931 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
4932 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
4933 // CHECK4-NEXT:  entry:
4934 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4935 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4936 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4937 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
4938 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4939 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4940 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4941 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
4942 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
4943 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
4944 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
4945 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
4946 // CHECK4-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
4947 // CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
4948 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
4949 // CHECK4-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
4950 // CHECK4-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
4951 // CHECK4-NEXT:    ret void
4952 //
4953 //
4954 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
4955 // 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:%.*]]) #[[ATTR2]] {
4956 // CHECK4-NEXT:  entry:
4957 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4958 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
4959 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4960 // CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
4961 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
4962 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
4963 // CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
4964 // CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
4965 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
4966 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
4967 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
4968 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
4969 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4970 // CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
4971 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
4972 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
4973 // CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
4974 // CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
4975 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
4976 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
4977 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4978 // CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
4979 // CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
4980 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
4981 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
4982 // CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
4983 // CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
4984 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
4985 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
4986 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
4987 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
4988 // CHECK4-NEXT:    ret void
4989 //
4990 //
4991 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..9
4992 // 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:%.*]]) #[[ATTR2]] {
4993 // CHECK4-NEXT:  entry:
4994 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4995 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4996 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
4997 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
4998 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
4999 // CHECK4-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
5000 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
5001 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5002 // CHECK4-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
5003 // CHECK4-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
5004 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
5005 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5006 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5007 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5008 // CHECK4-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
5009 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5010 // CHECK4-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
5011 // CHECK4-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
5012 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5013 // CHECK4-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
5014 // CHECK4-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
5015 // CHECK4-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
5016 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
5017 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5018 // CHECK4-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
5019 // CHECK4-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
5020 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5021 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
5022 // CHECK4-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
5023 // CHECK4-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
5024 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
5025 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
5026 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
5027 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
5028 // CHECK4-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
5029 // CHECK4-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
5030 // CHECK4-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
5031 // CHECK4-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
5032 // CHECK4-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
5033 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
5034 // CHECK4-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
5035 // CHECK4-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
5036 // CHECK4-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
5037 // CHECK4-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
5038 // CHECK4-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
5039 // CHECK4-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
5040 // CHECK4-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
5041 // CHECK4-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
5042 // CHECK4-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
5043 // CHECK4-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
5044 // CHECK4-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
5045 // CHECK4-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
5046 // CHECK4-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
5047 // CHECK4-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
5048 // CHECK4-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
5049 // CHECK4-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
5050 // CHECK4-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
5051 // CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
5052 // CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
5053 // CHECK4-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
5054 // CHECK4-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
5055 // CHECK4-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
5056 // CHECK4-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
5057 // CHECK4-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
5058 // CHECK4-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
5059 // CHECK4-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
5060 // CHECK4-NEXT:    ret void
5061 //
5062 //
5063 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
5064 // CHECK4-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
5065 // CHECK4-NEXT:  entry:
5066 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
5067 // CHECK4-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
5068 // CHECK4-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
5069 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
5070 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
5071 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
5072 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP1]])
5073 // CHECK4-NEXT:    ret void
5074 //
5075 //
5076 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..12
5077 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
5078 // CHECK4-NEXT:  entry:
5079 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5080 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5081 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
5082 // CHECK4-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
5083 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5084 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5085 // CHECK4-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
5086 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
5087 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
5088 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
5089 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP1]])
5090 // CHECK4-NEXT:    ret void
5091 //
5092 //
5093 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..13
5094 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
5095 // CHECK4-NEXT:  entry:
5096 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5097 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5098 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
5099 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5100 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5101 // CHECK4-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
5102 // CHECK4-NEXT:    ret void
5103 //
5104 //
5105 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
5106 // CHECK4-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
5107 // CHECK4-NEXT:  entry:
5108 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
5109 // CHECK4-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
5110 // CHECK4-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
5111 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
5112 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
5113 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
5114 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP1]])
5115 // CHECK4-NEXT:    ret void
5116 //
5117 //
5118 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..16
5119 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
5120 // CHECK4-NEXT:  entry:
5121 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5122 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5123 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
5124 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5125 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5126 // CHECK4-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
5127 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
5128 // CHECK4-NEXT:    ret void
5129 //
5130 //
5131 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..17
5132 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
5133 // CHECK4-NEXT:  entry:
5134 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5135 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5136 // CHECK4-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
5137 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5138 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5139 // CHECK4-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
5140 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
5141 // CHECK4-NEXT:    ret void
5142 //
5143 //
5144 // CHECK4-LABEL: define {{[^@]+}}@_Z6bazzzziPi
5145 // CHECK4-SAME: (i32 noundef [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
5146 // CHECK4-NEXT:  entry:
5147 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5148 // CHECK4-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 4
5149 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
5150 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
5151 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
5152 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5153 // CHECK4-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 4
5154 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
5155 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5156 // CHECK4-NEXT:    [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32*
5157 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[TMP2]], align 4
5158 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5159 // CHECK4-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
5160 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[TMP4]], align 4
5161 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5162 // CHECK4-NEXT:    store i8* null, i8** [[TMP5]], align 4
5163 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5164 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5165 // CHECK4-NEXT:    [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5166 // CHECK4-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
5167 // CHECK4-NEXT:    br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5168 // CHECK4:       omp_offload.failed:
5169 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i32 [[TMP0]]) #[[ATTR3]]
5170 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5171 // CHECK4:       omp_offload.cont:
5172 // CHECK4-NEXT:    ret void
5173 //
5174 //
5175 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
5176 // CHECK4-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
5177 // CHECK4-NEXT:  entry:
5178 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5179 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5180 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5181 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i32 [[TMP0]])
5182 // CHECK4-NEXT:    ret void
5183 //
5184 //
5185 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..20
5186 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
5187 // CHECK4-NEXT:  entry:
5188 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5189 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5190 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5191 // CHECK4-NEXT:    [[F:%.*]] = alloca i32*, align 4
5192 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5193 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5194 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5195 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5196 // CHECK4-NEXT:    ret void
5197 //
5198 //
5199 // CHECK4-LABEL: define {{[^@]+}}@_Z3bari
5200 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
5201 // CHECK4-NEXT:  entry:
5202 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5203 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
5204 // CHECK4-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
5205 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5206 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
5207 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
5208 // CHECK4-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
5209 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
5210 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
5211 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5212 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
5213 // CHECK4-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
5214 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
5215 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
5216 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
5217 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
5218 // CHECK4-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
5219 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5220 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
5221 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
5222 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
5223 // CHECK4-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
5224 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
5225 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
5226 // CHECK4-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
5227 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
5228 // CHECK4-NEXT:    ret i32 [[TMP8]]
5229 //
5230 //
5231 // CHECK4-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
5232 // CHECK4-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
5233 // CHECK4-NEXT:  entry:
5234 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5235 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5236 // CHECK4-NEXT:    [[B:%.*]] = alloca i32, align 4
5237 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
5238 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5239 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
5240 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
5241 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
5242 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
5243 // CHECK4-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
5244 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5245 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5246 // CHECK4-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5247 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
5248 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
5249 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
5250 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
5251 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
5252 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
5253 // CHECK4-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
5254 // CHECK4-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
5255 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
5256 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
5257 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
5258 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
5259 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
5260 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
5261 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5262 // CHECK4:       omp_if.then:
5263 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
5264 // CHECK4-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
5265 // CHECK4-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
5266 // CHECK4-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
5267 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
5268 // CHECK4-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i32 40, i1 false)
5269 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5270 // CHECK4-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
5271 // CHECK4-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
5272 // CHECK4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5273 // CHECK4-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
5274 // CHECK4-NEXT:    store double* [[A]], double** [[TMP14]], align 4
5275 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5276 // CHECK4-NEXT:    store i8* null, i8** [[TMP15]], align 4
5277 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
5278 // CHECK4-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
5279 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
5280 // CHECK4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
5281 // CHECK4-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
5282 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
5283 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
5284 // CHECK4-NEXT:    store i8* null, i8** [[TMP20]], align 4
5285 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
5286 // CHECK4-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
5287 // CHECK4-NEXT:    store i32 2, i32* [[TMP22]], align 4
5288 // CHECK4-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
5289 // CHECK4-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
5290 // CHECK4-NEXT:    store i32 2, i32* [[TMP24]], align 4
5291 // CHECK4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
5292 // CHECK4-NEXT:    store i8* null, i8** [[TMP25]], align 4
5293 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
5294 // CHECK4-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
5295 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
5296 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
5297 // CHECK4-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
5298 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
5299 // CHECK4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
5300 // CHECK4-NEXT:    store i8* null, i8** [[TMP30]], align 4
5301 // CHECK4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
5302 // CHECK4-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
5303 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
5304 // CHECK4-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
5305 // CHECK4-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
5306 // CHECK4-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
5307 // CHECK4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
5308 // CHECK4-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
5309 // CHECK4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
5310 // CHECK4-NEXT:    store i8* null, i8** [[TMP36]], align 4
5311 // CHECK4-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5312 // CHECK4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5313 // CHECK4-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
5314 // CHECK4-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5315 // CHECK4-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
5316 // CHECK4-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5317 // CHECK4:       omp_offload.failed:
5318 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
5319 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5320 // CHECK4:       omp_offload.cont:
5321 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5322 // CHECK4:       omp_if.else:
5323 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
5324 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5325 // CHECK4:       omp_if.end:
5326 // CHECK4-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
5327 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
5328 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
5329 // CHECK4-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
5330 // CHECK4-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
5331 // CHECK4-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
5332 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
5333 // CHECK4-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5334 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
5335 // CHECK4-NEXT:    ret i32 [[ADD3]]
5336 //
5337 //
5338 // CHECK4-LABEL: define {{[^@]+}}@_ZL7fstatici
5339 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
5340 // CHECK4-NEXT:  entry:
5341 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5342 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
5343 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
5344 // CHECK4-NEXT:    [[AAA:%.*]] = alloca i8, align 1
5345 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
5346 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5347 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5348 // CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
5349 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
5350 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
5351 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
5352 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5353 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
5354 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
5355 // CHECK4-NEXT:    store i8 0, i8* [[AAA]], align 1
5356 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
5357 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
5358 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
5359 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
5360 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5361 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
5362 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5363 // CHECK4-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
5364 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
5365 // CHECK4-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
5366 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
5367 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
5368 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
5369 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5370 // CHECK4:       omp_if.then:
5371 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5372 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
5373 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
5374 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5375 // CHECK4-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
5376 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
5377 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5378 // CHECK4-NEXT:    store i8* null, i8** [[TMP11]], align 4
5379 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
5380 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
5381 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
5382 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
5383 // CHECK4-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
5384 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
5385 // CHECK4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
5386 // CHECK4-NEXT:    store i8* null, i8** [[TMP16]], align 4
5387 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
5388 // CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
5389 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
5390 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
5391 // CHECK4-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
5392 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
5393 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
5394 // CHECK4-NEXT:    store i8* null, i8** [[TMP21]], align 4
5395 // CHECK4-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
5396 // CHECK4-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
5397 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
5398 // CHECK4-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
5399 // CHECK4-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
5400 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
5401 // CHECK4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
5402 // CHECK4-NEXT:    store i8* null, i8** [[TMP26]], align 4
5403 // CHECK4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5404 // CHECK4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5405 // CHECK4-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5406 // CHECK4-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
5407 // CHECK4-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5408 // CHECK4:       omp_offload.failed:
5409 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
5410 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5411 // CHECK4:       omp_offload.cont:
5412 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5413 // CHECK4:       omp_if.else:
5414 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
5415 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5416 // CHECK4:       omp_if.end:
5417 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
5418 // CHECK4-NEXT:    ret i32 [[TMP31]]
5419 //
5420 //
5421 // CHECK4-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
5422 // CHECK4-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
5423 // CHECK4-NEXT:  entry:
5424 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5425 // CHECK4-NEXT:    [[A:%.*]] = alloca i32, align 4
5426 // CHECK4-NEXT:    [[AA:%.*]] = alloca i16, align 2
5427 // CHECK4-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
5428 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5429 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5430 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
5431 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
5432 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
5433 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5434 // CHECK4-NEXT:    store i32 0, i32* [[A]], align 4
5435 // CHECK4-NEXT:    store i16 0, i16* [[AA]], align 2
5436 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
5437 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
5438 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
5439 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
5440 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5441 // CHECK4-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
5442 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5443 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
5444 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
5445 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5446 // CHECK4:       omp_if.then:
5447 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5448 // CHECK4-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
5449 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
5450 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5451 // CHECK4-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
5452 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
5453 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5454 // CHECK4-NEXT:    store i8* null, i8** [[TMP9]], align 4
5455 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
5456 // CHECK4-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
5457 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
5458 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
5459 // CHECK4-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
5460 // CHECK4-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
5461 // CHECK4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
5462 // CHECK4-NEXT:    store i8* null, i8** [[TMP14]], align 4
5463 // CHECK4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
5464 // CHECK4-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
5465 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
5466 // CHECK4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
5467 // CHECK4-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
5468 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
5469 // CHECK4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
5470 // CHECK4-NEXT:    store i8* null, i8** [[TMP19]], align 4
5471 // CHECK4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5472 // CHECK4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5473 // CHECK4-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5474 // CHECK4-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
5475 // CHECK4-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5476 // CHECK4:       omp_offload.failed:
5477 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
5478 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5479 // CHECK4:       omp_offload.cont:
5480 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5481 // CHECK4:       omp_if.else:
5482 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
5483 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5484 // CHECK4:       omp_if.end:
5485 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
5486 // CHECK4-NEXT:    ret i32 [[TMP24]]
5487 //
5488 //
5489 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
5490 // CHECK4-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
5491 // CHECK4-NEXT:  entry:
5492 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5493 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5494 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5495 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5496 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
5497 // CHECK4-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
5498 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5499 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5500 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5501 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5502 // CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
5503 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5504 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5505 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5506 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
5507 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
5508 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
5509 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
5510 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
5511 // CHECK4-NEXT:    ret void
5512 //
5513 //
5514 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..23
5515 // 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]] {
5516 // CHECK4-NEXT:  entry:
5517 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5518 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5519 // CHECK4-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
5520 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
5521 // CHECK4-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5522 // CHECK4-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5523 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
5524 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5525 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5526 // CHECK4-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
5527 // CHECK4-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
5528 // CHECK4-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5529 // CHECK4-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5530 // CHECK4-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
5531 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
5532 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5533 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5534 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
5535 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
5536 // CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
5537 // CHECK4-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
5538 // CHECK4-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
5539 // CHECK4-NEXT:    store double [[ADD]], double* [[A]], align 4
5540 // CHECK4-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
5541 // CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
5542 // CHECK4-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
5543 // CHECK4-NEXT:    store double [[INC]], double* [[A3]], align 4
5544 // CHECK4-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
5545 // CHECK4-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
5546 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
5547 // CHECK4-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
5548 // CHECK4-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
5549 // CHECK4-NEXT:    ret void
5550 //
5551 //
5552 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
5553 // CHECK4-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5554 // CHECK4-NEXT:  entry:
5555 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5556 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5557 // CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
5558 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5559 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5560 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5561 // CHECK4-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
5562 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5563 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5564 // CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
5565 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5566 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5567 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
5568 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5569 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5570 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
5571 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
5572 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
5573 // CHECK4-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5574 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
5575 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5576 // CHECK4-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
5577 // CHECK4-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
5578 // CHECK4-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
5579 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
5580 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
5581 // CHECK4-NEXT:    ret void
5582 //
5583 //
5584 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..26
5585 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5586 // CHECK4-NEXT:  entry:
5587 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5588 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5589 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5590 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5591 // CHECK4-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
5592 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5593 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5594 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5595 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5596 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5597 // CHECK4-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
5598 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5599 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5600 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
5601 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5602 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5603 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
5604 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
5605 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
5606 // CHECK4-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
5607 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
5608 // CHECK4-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
5609 // CHECK4-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
5610 // CHECK4-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
5611 // CHECK4-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
5612 // CHECK4-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
5613 // CHECK4-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
5614 // CHECK4-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
5615 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
5616 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5617 // CHECK4-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
5618 // CHECK4-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
5619 // CHECK4-NEXT:    ret void
5620 //
5621 //
5622 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
5623 // CHECK4-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
5624 // CHECK4-NEXT:  entry:
5625 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5626 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5627 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5628 // CHECK4-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5629 // CHECK4-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5630 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5631 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5632 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5633 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5634 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5635 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5636 // CHECK4-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
5637 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
5638 // CHECK4-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
5639 // CHECK4-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5640 // CHECK4-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
5641 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5642 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
5643 // CHECK4-NEXT:    ret void
5644 //
5645 //
5646 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..29
5647 // 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]] {
5648 // CHECK4-NEXT:  entry:
5649 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5650 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5651 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5652 // CHECK4-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5653 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
5654 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5655 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5656 // CHECK4-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5657 // CHECK4-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5658 // CHECK4-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
5659 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5660 // CHECK4-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
5661 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5662 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
5663 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
5664 // CHECK4-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
5665 // CHECK4-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
5666 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
5667 // CHECK4-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
5668 // CHECK4-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
5669 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
5670 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5671 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
5672 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
5673 // CHECK4-NEXT:    ret void
5674 //
5675 //
5676 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5677 // CHECK4-SAME: () #[[ATTR4]] {
5678 // CHECK4-NEXT:  entry:
5679 // CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
5680 // CHECK4-NEXT:    ret void
5681 //
5682 //
5683 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
5684 // CHECK9-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
5685 // CHECK9-NEXT:  entry:
5686 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5687 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5688 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
5689 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
5690 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
5691 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5692 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5693 // CHECK9-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
5694 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5695 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5696 // CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
5697 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
5698 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
5699 // CHECK9-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
5700 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
5701 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
5702 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
5703 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
5704 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
5705 // CHECK9-NEXT:    ret void
5706 //
5707 //
5708 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
5709 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
5710 // CHECK9-NEXT:  entry:
5711 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5712 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5713 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5714 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5715 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5716 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5717 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5718 // CHECK9-NEXT:    ret void
5719 //
5720 //
5721 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
5722 // CHECK9-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
5723 // CHECK9-NEXT:  entry:
5724 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5725 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
5726 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5727 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5728 // CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
5729 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
5730 // CHECK9-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
5731 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
5732 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
5733 // CHECK9-NEXT:    ret void
5734 //
5735 //
5736 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
5737 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
5738 // CHECK9-NEXT:  entry:
5739 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5740 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5741 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5742 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5743 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5744 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5745 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5746 // CHECK9-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
5747 // CHECK9-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
5748 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
5749 // CHECK9-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
5750 // CHECK9-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
5751 // CHECK9-NEXT:    ret void
5752 //
5753 //
5754 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
5755 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
5756 // CHECK9-NEXT:  entry:
5757 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5758 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5759 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
5760 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
5761 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5762 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5763 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5764 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5765 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
5766 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
5767 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
5768 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
5769 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
5770 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
5771 // CHECK9-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
5772 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
5773 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
5774 // CHECK9-NEXT:    ret void
5775 //
5776 //
5777 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
5778 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
5779 // CHECK9-NEXT:  entry:
5780 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5781 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5782 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5783 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
5784 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5785 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5786 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5787 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
5788 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5789 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
5790 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
5791 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
5792 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
5793 // CHECK9-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
5794 // CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
5795 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
5796 // CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
5797 // CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
5798 // CHECK9-NEXT:    ret void
5799 //
5800 //
5801 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
5802 // 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:%.*]]) #[[ATTR0]] {
5803 // CHECK9-NEXT:  entry:
5804 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5805 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
5806 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
5807 // CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
5808 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
5809 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
5810 // CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
5811 // CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
5812 // CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
5813 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
5814 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5815 // CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
5816 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5817 // CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
5818 // CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
5819 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
5820 // CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
5821 // CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
5822 // CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
5823 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5824 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
5825 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5826 // CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
5827 // CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
5828 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
5829 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
5830 // CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
5831 // CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
5832 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
5833 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
5834 // CHECK9-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
5835 // CHECK9-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
5836 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
5837 // CHECK9-NEXT:    ret void
5838 //
5839 //
5840 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
5841 // 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:%.*]]) #[[ATTR0]] {
5842 // CHECK9-NEXT:  entry:
5843 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5844 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5845 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
5846 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
5847 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
5848 // CHECK9-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
5849 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
5850 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
5851 // CHECK9-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
5852 // CHECK9-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
5853 // CHECK9-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
5854 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5855 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5856 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
5857 // CHECK9-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
5858 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5859 // CHECK9-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
5860 // CHECK9-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
5861 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
5862 // CHECK9-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
5863 // CHECK9-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
5864 // CHECK9-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
5865 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
5866 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
5867 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5868 // CHECK9-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
5869 // CHECK9-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
5870 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
5871 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
5872 // CHECK9-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
5873 // CHECK9-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
5874 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
5875 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
5876 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
5877 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
5878 // CHECK9-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
5879 // CHECK9-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
5880 // CHECK9-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
5881 // CHECK9-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
5882 // CHECK9-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
5883 // CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
5884 // CHECK9-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
5885 // CHECK9-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
5886 // CHECK9-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
5887 // CHECK9-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
5888 // CHECK9-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
5889 // CHECK9-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
5890 // CHECK9-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
5891 // CHECK9-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
5892 // CHECK9-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
5893 // CHECK9-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
5894 // CHECK9-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
5895 // CHECK9-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
5896 // CHECK9-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
5897 // CHECK9-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
5898 // CHECK9-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
5899 // CHECK9-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
5900 // CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
5901 // CHECK9-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
5902 // CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
5903 // CHECK9-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
5904 // CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
5905 // CHECK9-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
5906 // CHECK9-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
5907 // CHECK9-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
5908 // CHECK9-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
5909 // CHECK9-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
5910 // CHECK9-NEXT:    ret void
5911 //
5912 //
5913 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
5914 // CHECK9-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
5915 // CHECK9-NEXT:  entry:
5916 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
5917 // CHECK9-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
5918 // CHECK9-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
5919 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
5920 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
5921 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
5922 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
5923 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
5924 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]])
5925 // CHECK9-NEXT:    ret void
5926 //
5927 //
5928 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
5929 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
5930 // CHECK9-NEXT:  entry:
5931 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5932 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5933 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
5934 // CHECK9-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
5935 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5936 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5937 // CHECK9-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
5938 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
5939 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
5940 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
5941 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
5942 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
5943 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP1]])
5944 // CHECK9-NEXT:    ret void
5945 //
5946 //
5947 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
5948 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
5949 // CHECK9-NEXT:  entry:
5950 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5951 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5952 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
5953 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5954 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5955 // CHECK9-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
5956 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
5957 // CHECK9-NEXT:    ret void
5958 //
5959 //
5960 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
5961 // CHECK9-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
5962 // CHECK9-NEXT:  entry:
5963 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
5964 // CHECK9-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
5965 // CHECK9-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
5966 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
5967 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
5968 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
5969 // CHECK9-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
5970 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
5971 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]])
5972 // CHECK9-NEXT:    ret void
5973 //
5974 //
5975 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
5976 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
5977 // CHECK9-NEXT:  entry:
5978 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5979 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5980 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
5981 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5982 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5983 // CHECK9-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
5984 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
5985 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[CONV]])
5986 // CHECK9-NEXT:    ret void
5987 //
5988 //
5989 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
5990 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
5991 // CHECK9-NEXT:  entry:
5992 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5993 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5994 // CHECK9-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
5995 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5996 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5997 // CHECK9-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
5998 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
5999 // CHECK9-NEXT:    ret void
6000 //
6001 //
6002 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
6003 // CHECK9-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
6004 // CHECK9-NEXT:  entry:
6005 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6006 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6007 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6008 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i64 [[TMP0]])
6009 // CHECK9-NEXT:    ret void
6010 //
6011 //
6012 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..8
6013 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
6014 // CHECK9-NEXT:  entry:
6015 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6016 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6017 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6018 // CHECK9-NEXT:    [[F:%.*]] = alloca i32*, align 8
6019 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6020 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6021 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6022 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6023 // CHECK9-NEXT:    ret void
6024 //
6025 //
6026 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
6027 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6028 // CHECK9-NEXT:  entry:
6029 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6030 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6031 // CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
6032 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6033 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6034 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6035 // CHECK9-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
6036 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6037 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6038 // CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
6039 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6040 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6041 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6042 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
6043 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6044 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6045 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6046 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
6047 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
6048 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
6049 // CHECK9-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6050 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
6051 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6052 // CHECK9-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
6053 // CHECK9-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
6054 // CHECK9-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
6055 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
6056 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
6057 // CHECK9-NEXT:    ret void
6058 //
6059 //
6060 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..9
6061 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6062 // CHECK9-NEXT:  entry:
6063 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6064 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6065 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6066 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6067 // CHECK9-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
6068 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6069 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6070 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6071 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6072 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6073 // CHECK9-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
6074 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6075 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6076 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6077 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
6078 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6079 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6080 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
6081 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6082 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
6083 // CHECK9-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
6084 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6085 // CHECK9-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
6086 // CHECK9-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
6087 // CHECK9-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
6088 // CHECK9-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
6089 // CHECK9-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
6090 // CHECK9-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
6091 // CHECK9-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
6092 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
6093 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6094 // CHECK9-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
6095 // CHECK9-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
6096 // CHECK9-NEXT:    ret void
6097 //
6098 //
6099 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
6100 // CHECK9-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
6101 // CHECK9-NEXT:  entry:
6102 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6103 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
6104 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6105 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6106 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
6107 // CHECK9-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
6108 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
6109 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
6110 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6111 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6112 // CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
6113 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
6114 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
6115 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6116 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6117 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
6118 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
6119 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
6120 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
6121 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
6122 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
6123 // CHECK9-NEXT:    ret void
6124 //
6125 //
6126 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
6127 // 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]] {
6128 // CHECK9-NEXT:  entry:
6129 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6130 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6131 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6132 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
6133 // CHECK9-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6134 // CHECK9-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6135 // CHECK9-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
6136 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6137 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6138 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
6139 // CHECK9-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
6140 // CHECK9-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6141 // CHECK9-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6142 // CHECK9-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
6143 // CHECK9-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
6144 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
6145 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6146 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6147 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
6148 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
6149 // CHECK9-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
6150 // CHECK9-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
6151 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
6152 // CHECK9-NEXT:    store double [[ADD]], double* [[A]], align 8
6153 // CHECK9-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6154 // CHECK9-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
6155 // CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
6156 // CHECK9-NEXT:    store double [[INC]], double* [[A4]], align 8
6157 // CHECK9-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
6158 // CHECK9-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
6159 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
6160 // CHECK9-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
6161 // CHECK9-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
6162 // CHECK9-NEXT:    ret void
6163 //
6164 //
6165 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
6166 // CHECK9-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6167 // CHECK9-NEXT:  entry:
6168 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6169 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6170 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6171 // CHECK9-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6172 // CHECK9-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6173 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6174 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6175 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6176 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6177 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6178 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6179 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6180 // CHECK9-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6181 // CHECK9-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
6182 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
6183 // CHECK9-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
6184 // CHECK9-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6185 // CHECK9-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
6186 // CHECK9-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6187 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
6188 // CHECK9-NEXT:    ret void
6189 //
6190 //
6191 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
6192 // 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]] {
6193 // CHECK9-NEXT:  entry:
6194 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6195 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6196 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6197 // CHECK9-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6198 // CHECK9-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6199 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6200 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6201 // CHECK9-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6202 // CHECK9-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6203 // CHECK9-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6204 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6205 // CHECK9-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6206 // CHECK9-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6207 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6208 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
6209 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6210 // CHECK9-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
6211 // CHECK9-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
6212 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
6213 // CHECK9-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
6214 // CHECK9-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
6215 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
6216 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6217 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
6218 // CHECK9-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
6219 // CHECK9-NEXT:    ret void
6220 //
6221 //
6222 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
6223 // CHECK10-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
6224 // CHECK10-NEXT:  entry:
6225 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6226 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
6227 // CHECK10-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
6228 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6229 // CHECK10-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
6230 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6231 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
6232 // CHECK10-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
6233 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6234 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
6235 // CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
6236 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
6237 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
6238 // CHECK10-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
6239 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
6240 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6241 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
6242 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6243 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
6244 // CHECK10-NEXT:    ret void
6245 //
6246 //
6247 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
6248 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
6249 // CHECK10-NEXT:  entry:
6250 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6251 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6252 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6253 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6254 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6255 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6256 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6257 // CHECK10-NEXT:    ret void
6258 //
6259 //
6260 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
6261 // CHECK10-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
6262 // CHECK10-NEXT:  entry:
6263 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6264 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6265 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6266 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6267 // CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
6268 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6269 // CHECK10-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
6270 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6271 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
6272 // CHECK10-NEXT:    ret void
6273 //
6274 //
6275 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
6276 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
6277 // CHECK10-NEXT:  entry:
6278 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6279 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6280 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6281 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6282 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6283 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6284 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6285 // CHECK10-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
6286 // CHECK10-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
6287 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
6288 // CHECK10-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
6289 // CHECK10-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
6290 // CHECK10-NEXT:    ret void
6291 //
6292 //
6293 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
6294 // CHECK10-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
6295 // CHECK10-NEXT:  entry:
6296 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6297 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6298 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6299 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6300 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6301 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6302 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6303 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6304 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
6305 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6306 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
6307 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
6308 // CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
6309 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6310 // CHECK10-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
6311 // CHECK10-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6312 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
6313 // CHECK10-NEXT:    ret void
6314 //
6315 //
6316 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
6317 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
6318 // CHECK10-NEXT:  entry:
6319 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6320 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6321 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6322 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6323 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6324 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6325 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6326 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6327 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6328 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6329 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
6330 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
6331 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6332 // CHECK10-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
6333 // CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
6334 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
6335 // CHECK10-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
6336 // CHECK10-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
6337 // CHECK10-NEXT:    ret void
6338 //
6339 //
6340 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
6341 // 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:%.*]]) #[[ATTR0]] {
6342 // CHECK10-NEXT:  entry:
6343 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6344 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
6345 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6346 // CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
6347 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
6348 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6349 // CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
6350 // CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
6351 // CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
6352 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6353 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6354 // CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
6355 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6356 // CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
6357 // CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
6358 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6359 // CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
6360 // CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
6361 // CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
6362 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6363 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
6364 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6365 // CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
6366 // CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
6367 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6368 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
6369 // CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
6370 // CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
6371 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
6372 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6373 // CHECK10-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
6374 // CHECK10-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
6375 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
6376 // CHECK10-NEXT:    ret void
6377 //
6378 //
6379 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
6380 // 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:%.*]]) #[[ATTR0]] {
6381 // CHECK10-NEXT:  entry:
6382 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6383 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6384 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6385 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
6386 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6387 // CHECK10-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
6388 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
6389 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6390 // CHECK10-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
6391 // CHECK10-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
6392 // CHECK10-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
6393 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6394 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6395 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6396 // CHECK10-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
6397 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6398 // CHECK10-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
6399 // CHECK10-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
6400 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6401 // CHECK10-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
6402 // CHECK10-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
6403 // CHECK10-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
6404 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6405 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
6406 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6407 // CHECK10-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
6408 // CHECK10-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
6409 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6410 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
6411 // CHECK10-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
6412 // CHECK10-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
6413 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
6414 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
6415 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6416 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
6417 // CHECK10-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
6418 // CHECK10-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
6419 // CHECK10-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
6420 // CHECK10-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
6421 // CHECK10-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
6422 // CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
6423 // CHECK10-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
6424 // CHECK10-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
6425 // CHECK10-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
6426 // CHECK10-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
6427 // CHECK10-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
6428 // CHECK10-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
6429 // CHECK10-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
6430 // CHECK10-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
6431 // CHECK10-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
6432 // CHECK10-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
6433 // CHECK10-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
6434 // CHECK10-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
6435 // CHECK10-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
6436 // CHECK10-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
6437 // CHECK10-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
6438 // CHECK10-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
6439 // CHECK10-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
6440 // CHECK10-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
6441 // CHECK10-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
6442 // CHECK10-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
6443 // CHECK10-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
6444 // CHECK10-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
6445 // CHECK10-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
6446 // CHECK10-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
6447 // CHECK10-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
6448 // CHECK10-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
6449 // CHECK10-NEXT:    ret void
6450 //
6451 //
6452 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
6453 // CHECK10-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
6454 // CHECK10-NEXT:  entry:
6455 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
6456 // CHECK10-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
6457 // CHECK10-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
6458 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
6459 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
6460 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
6461 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
6462 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
6463 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]])
6464 // CHECK10-NEXT:    ret void
6465 //
6466 //
6467 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
6468 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
6469 // CHECK10-NEXT:  entry:
6470 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6471 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6472 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
6473 // CHECK10-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
6474 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6475 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6476 // CHECK10-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
6477 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
6478 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
6479 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
6480 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
6481 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
6482 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP1]])
6483 // CHECK10-NEXT:    ret void
6484 //
6485 //
6486 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
6487 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
6488 // CHECK10-NEXT:  entry:
6489 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6490 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6491 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
6492 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6493 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6494 // CHECK10-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
6495 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
6496 // CHECK10-NEXT:    ret void
6497 //
6498 //
6499 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
6500 // CHECK10-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
6501 // CHECK10-NEXT:  entry:
6502 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
6503 // CHECK10-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
6504 // CHECK10-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
6505 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
6506 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
6507 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
6508 // CHECK10-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
6509 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
6510 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]])
6511 // CHECK10-NEXT:    ret void
6512 //
6513 //
6514 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
6515 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
6516 // CHECK10-NEXT:  entry:
6517 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6518 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6519 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
6520 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6521 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6522 // CHECK10-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
6523 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
6524 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[CONV]])
6525 // CHECK10-NEXT:    ret void
6526 //
6527 //
6528 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..7
6529 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
6530 // CHECK10-NEXT:  entry:
6531 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6532 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6533 // CHECK10-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
6534 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6535 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6536 // CHECK10-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
6537 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
6538 // CHECK10-NEXT:    ret void
6539 //
6540 //
6541 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
6542 // CHECK10-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
6543 // CHECK10-NEXT:  entry:
6544 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6545 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6546 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6547 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i64 [[TMP0]])
6548 // CHECK10-NEXT:    ret void
6549 //
6550 //
6551 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..8
6552 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
6553 // CHECK10-NEXT:  entry:
6554 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6555 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6556 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6557 // CHECK10-NEXT:    [[F:%.*]] = alloca i32*, align 8
6558 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6559 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6560 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6561 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6562 // CHECK10-NEXT:    ret void
6563 //
6564 //
6565 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
6566 // CHECK10-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6567 // CHECK10-NEXT:  entry:
6568 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6569 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6570 // CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
6571 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6572 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6573 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6574 // CHECK10-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
6575 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6576 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6577 // CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
6578 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6579 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6580 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6581 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
6582 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6583 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6584 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6585 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
6586 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
6587 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
6588 // CHECK10-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6589 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
6590 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6591 // CHECK10-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
6592 // CHECK10-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
6593 // CHECK10-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
6594 // CHECK10-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
6595 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
6596 // CHECK10-NEXT:    ret void
6597 //
6598 //
6599 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..9
6600 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6601 // CHECK10-NEXT:  entry:
6602 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6603 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6604 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6605 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6606 // CHECK10-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
6607 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6608 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6609 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6610 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6611 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6612 // CHECK10-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
6613 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6614 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6615 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6616 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
6617 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6618 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6619 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
6620 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6621 // CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
6622 // CHECK10-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
6623 // CHECK10-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6624 // CHECK10-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
6625 // CHECK10-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
6626 // CHECK10-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
6627 // CHECK10-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
6628 // CHECK10-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
6629 // CHECK10-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
6630 // CHECK10-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
6631 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
6632 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6633 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
6634 // CHECK10-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
6635 // CHECK10-NEXT:    ret void
6636 //
6637 //
6638 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
6639 // CHECK10-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
6640 // CHECK10-NEXT:  entry:
6641 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6642 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
6643 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6644 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6645 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
6646 // CHECK10-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
6647 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
6648 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
6649 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6650 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6651 // CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
6652 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
6653 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
6654 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6655 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6656 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
6657 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
6658 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
6659 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
6660 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
6661 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
6662 // CHECK10-NEXT:    ret void
6663 //
6664 //
6665 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..10
6666 // 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]] {
6667 // CHECK10-NEXT:  entry:
6668 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6669 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6670 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6671 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
6672 // CHECK10-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
6673 // CHECK10-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
6674 // CHECK10-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
6675 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6676 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6677 // CHECK10-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
6678 // CHECK10-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
6679 // CHECK10-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
6680 // CHECK10-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
6681 // CHECK10-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
6682 // CHECK10-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
6683 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
6684 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
6685 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
6686 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
6687 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
6688 // CHECK10-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
6689 // CHECK10-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
6690 // CHECK10-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
6691 // CHECK10-NEXT:    store double [[ADD]], double* [[A]], align 8
6692 // CHECK10-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6693 // CHECK10-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
6694 // CHECK10-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
6695 // CHECK10-NEXT:    store double [[INC]], double* [[A4]], align 8
6696 // CHECK10-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
6697 // CHECK10-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
6698 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
6699 // CHECK10-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
6700 // CHECK10-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
6701 // CHECK10-NEXT:    ret void
6702 //
6703 //
6704 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
6705 // CHECK10-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
6706 // CHECK10-NEXT:  entry:
6707 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6708 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6709 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6710 // CHECK10-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
6711 // CHECK10-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
6712 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6713 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6714 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6715 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6716 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6717 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6718 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6719 // CHECK10-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
6720 // CHECK10-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
6721 // CHECK10-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
6722 // CHECK10-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
6723 // CHECK10-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
6724 // CHECK10-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
6725 // CHECK10-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
6726 // CHECK10-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
6727 // CHECK10-NEXT:    ret void
6728 //
6729 //
6730 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..11
6731 // 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]] {
6732 // CHECK10-NEXT:  entry:
6733 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6734 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6735 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
6736 // CHECK10-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
6737 // CHECK10-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
6738 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6739 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6740 // CHECK10-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
6741 // CHECK10-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
6742 // CHECK10-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
6743 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
6744 // CHECK10-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
6745 // CHECK10-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
6746 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
6747 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
6748 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
6749 // CHECK10-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
6750 // CHECK10-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
6751 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
6752 // CHECK10-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
6753 // CHECK10-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
6754 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
6755 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6756 // CHECK10-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
6757 // CHECK10-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
6758 // CHECK10-NEXT:    ret void
6759 //
6760 //
6761 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
6762 // CHECK11-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
6763 // CHECK11-NEXT:  entry:
6764 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6765 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
6766 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
6767 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6768 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
6769 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6770 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6771 // CHECK11-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
6772 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6773 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6774 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
6775 // CHECK11-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
6776 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
6777 // CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6778 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
6779 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6780 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
6781 // CHECK11-NEXT:    ret void
6782 //
6783 //
6784 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
6785 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
6786 // CHECK11-NEXT:  entry:
6787 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6788 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6789 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6790 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6791 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6792 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6793 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6794 // CHECK11-NEXT:    ret void
6795 //
6796 //
6797 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
6798 // CHECK11-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
6799 // CHECK11-NEXT:  entry:
6800 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6801 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6802 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6803 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6804 // CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
6805 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6806 // CHECK11-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
6807 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6808 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
6809 // CHECK11-NEXT:    ret void
6810 //
6811 //
6812 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
6813 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
6814 // CHECK11-NEXT:  entry:
6815 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6816 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6817 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6818 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6819 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6820 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6821 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6822 // CHECK11-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
6823 // CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
6824 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
6825 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
6826 // CHECK11-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
6827 // CHECK11-NEXT:    ret void
6828 //
6829 //
6830 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
6831 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
6832 // CHECK11-NEXT:  entry:
6833 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6834 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6835 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6836 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6837 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6838 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6839 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6840 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
6841 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
6842 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
6843 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
6844 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6845 // CHECK11-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
6846 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6847 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
6848 // CHECK11-NEXT:    ret void
6849 //
6850 //
6851 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
6852 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
6853 // CHECK11-NEXT:  entry:
6854 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6855 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6856 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6857 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6858 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6859 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6860 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6861 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6862 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6863 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
6864 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
6865 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
6866 // CHECK11-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
6867 // CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
6868 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
6869 // CHECK11-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
6870 // CHECK11-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
6871 // CHECK11-NEXT:    ret void
6872 //
6873 //
6874 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
6875 // 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:%.*]]) #[[ATTR0]] {
6876 // CHECK11-NEXT:  entry:
6877 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6878 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
6879 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6880 // CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
6881 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
6882 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6883 // CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
6884 // CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
6885 // CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
6886 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6887 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6888 // CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
6889 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6890 // CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
6891 // CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
6892 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6893 // CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
6894 // CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
6895 // CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
6896 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
6897 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6898 // CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
6899 // CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
6900 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6901 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
6902 // CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
6903 // CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
6904 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
6905 // CHECK11-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
6906 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
6907 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
6908 // CHECK11-NEXT:    ret void
6909 //
6910 //
6911 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
6912 // 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:%.*]]) #[[ATTR0]] {
6913 // CHECK11-NEXT:  entry:
6914 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6915 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6916 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6917 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
6918 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6919 // CHECK11-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
6920 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
6921 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6922 // CHECK11-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
6923 // CHECK11-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
6924 // CHECK11-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
6925 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6926 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6927 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6928 // CHECK11-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
6929 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6930 // CHECK11-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
6931 // CHECK11-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
6932 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6933 // CHECK11-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
6934 // CHECK11-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
6935 // CHECK11-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
6936 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
6937 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6938 // CHECK11-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
6939 // CHECK11-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
6940 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6941 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
6942 // CHECK11-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
6943 // CHECK11-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
6944 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
6945 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
6946 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
6947 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
6948 // CHECK11-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
6949 // CHECK11-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
6950 // CHECK11-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
6951 // CHECK11-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
6952 // CHECK11-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
6953 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
6954 // CHECK11-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
6955 // CHECK11-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
6956 // CHECK11-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
6957 // CHECK11-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
6958 // CHECK11-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
6959 // CHECK11-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
6960 // CHECK11-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
6961 // CHECK11-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
6962 // CHECK11-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
6963 // CHECK11-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
6964 // CHECK11-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
6965 // CHECK11-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
6966 // CHECK11-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
6967 // CHECK11-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
6968 // CHECK11-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
6969 // CHECK11-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
6970 // CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
6971 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
6972 // CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
6973 // CHECK11-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
6974 // CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
6975 // CHECK11-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
6976 // CHECK11-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
6977 // CHECK11-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
6978 // CHECK11-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
6979 // CHECK11-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
6980 // CHECK11-NEXT:    ret void
6981 //
6982 //
6983 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
6984 // CHECK11-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
6985 // CHECK11-NEXT:  entry:
6986 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
6987 // CHECK11-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
6988 // CHECK11-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
6989 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
6990 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
6991 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
6992 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]])
6993 // CHECK11-NEXT:    ret void
6994 //
6995 //
6996 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
6997 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
6998 // CHECK11-NEXT:  entry:
6999 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7000 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7001 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7002 // CHECK11-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
7003 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7004 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7005 // CHECK11-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7006 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
7007 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
7008 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
7009 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7010 // CHECK11-NEXT:    ret void
7011 //
7012 //
7013 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
7014 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7015 // CHECK11-NEXT:  entry:
7016 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7017 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7018 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7019 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7020 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7021 // CHECK11-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7022 // CHECK11-NEXT:    ret void
7023 //
7024 //
7025 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
7026 // CHECK11-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7027 // CHECK11-NEXT:  entry:
7028 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7029 // CHECK11-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
7030 // CHECK11-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7031 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
7032 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
7033 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
7034 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7035 // CHECK11-NEXT:    ret void
7036 //
7037 //
7038 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
7039 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7040 // CHECK11-NEXT:  entry:
7041 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7042 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7043 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7044 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7045 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7046 // CHECK11-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7047 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
7048 // CHECK11-NEXT:    ret void
7049 //
7050 //
7051 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
7052 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
7053 // CHECK11-NEXT:  entry:
7054 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7055 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7056 // CHECK11-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
7057 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7058 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7059 // CHECK11-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
7060 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
7061 // CHECK11-NEXT:    ret void
7062 //
7063 //
7064 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
7065 // CHECK11-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
7066 // CHECK11-NEXT:  entry:
7067 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7068 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7069 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7070 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32 [[TMP0]])
7071 // CHECK11-NEXT:    ret void
7072 //
7073 //
7074 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..8
7075 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
7076 // CHECK11-NEXT:  entry:
7077 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7078 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7079 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7080 // CHECK11-NEXT:    [[F:%.*]] = alloca i32*, align 4
7081 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7082 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7083 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7084 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7085 // CHECK11-NEXT:    ret void
7086 //
7087 //
7088 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
7089 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7090 // CHECK11-NEXT:  entry:
7091 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7092 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7093 // CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
7094 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7095 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7096 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7097 // CHECK11-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
7098 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7099 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7100 // CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
7101 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7102 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7103 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
7104 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7105 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7106 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
7107 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
7108 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7109 // CHECK11-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7110 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
7111 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7112 // CHECK11-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
7113 // CHECK11-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
7114 // CHECK11-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
7115 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
7116 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
7117 // CHECK11-NEXT:    ret void
7118 //
7119 //
7120 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..9
7121 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7122 // CHECK11-NEXT:  entry:
7123 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7124 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7125 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7126 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7127 // CHECK11-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
7128 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7129 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7130 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7131 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7132 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7133 // CHECK11-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
7134 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7135 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7136 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
7137 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7138 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7139 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
7140 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7141 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
7142 // CHECK11-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
7143 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
7144 // CHECK11-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
7145 // CHECK11-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
7146 // CHECK11-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
7147 // CHECK11-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
7148 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
7149 // CHECK11-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
7150 // CHECK11-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
7151 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
7152 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7153 // CHECK11-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
7154 // CHECK11-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
7155 // CHECK11-NEXT:    ret void
7156 //
7157 //
7158 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
7159 // CHECK11-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
7160 // CHECK11-NEXT:  entry:
7161 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
7162 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
7163 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7164 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7165 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
7166 // CHECK11-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
7167 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
7168 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
7169 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7170 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7171 // CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
7172 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
7173 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7174 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7175 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
7176 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
7177 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
7178 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
7179 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
7180 // CHECK11-NEXT:    ret void
7181 //
7182 //
7183 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
7184 // 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]] {
7185 // CHECK11-NEXT:  entry:
7186 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7187 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7188 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
7189 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
7190 // CHECK11-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7191 // CHECK11-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7192 // CHECK11-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
7193 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7194 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7195 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
7196 // CHECK11-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
7197 // CHECK11-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7198 // CHECK11-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7199 // CHECK11-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
7200 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
7201 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7202 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7203 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
7204 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
7205 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
7206 // CHECK11-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
7207 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
7208 // CHECK11-NEXT:    store double [[ADD]], double* [[A]], align 4
7209 // CHECK11-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
7210 // CHECK11-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
7211 // CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
7212 // CHECK11-NEXT:    store double [[INC]], double* [[A3]], align 4
7213 // CHECK11-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
7214 // CHECK11-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
7215 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
7216 // CHECK11-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
7217 // CHECK11-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
7218 // CHECK11-NEXT:    ret void
7219 //
7220 //
7221 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
7222 // CHECK11-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7223 // CHECK11-NEXT:  entry:
7224 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7225 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7226 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7227 // CHECK11-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7228 // CHECK11-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7229 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7230 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7231 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7232 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7233 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7234 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7235 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
7236 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
7237 // CHECK11-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7238 // CHECK11-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7239 // CHECK11-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
7240 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7241 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
7242 // CHECK11-NEXT:    ret void
7243 //
7244 //
7245 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
7246 // 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]] {
7247 // CHECK11-NEXT:  entry:
7248 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7249 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7250 // CHECK11-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7251 // CHECK11-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7252 // CHECK11-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7253 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7254 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7255 // CHECK11-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7256 // CHECK11-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7257 // CHECK11-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7258 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7259 // CHECK11-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7260 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7261 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
7262 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7263 // CHECK11-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
7264 // CHECK11-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
7265 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
7266 // CHECK11-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
7267 // CHECK11-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
7268 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
7269 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7270 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
7271 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
7272 // CHECK11-NEXT:    ret void
7273 //
7274 //
7275 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
7276 // CHECK12-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
7277 // CHECK12-NEXT:  entry:
7278 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7279 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
7280 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
7281 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7282 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
7283 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7284 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7285 // CHECK12-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
7286 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7287 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7288 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
7289 // CHECK12-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
7290 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7291 // CHECK12-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7292 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
7293 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7294 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
7295 // CHECK12-NEXT:    ret void
7296 //
7297 //
7298 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
7299 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
7300 // CHECK12-NEXT:  entry:
7301 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7302 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7303 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7304 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7305 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7306 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7307 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7308 // CHECK12-NEXT:    ret void
7309 //
7310 //
7311 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
7312 // CHECK12-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
7313 // CHECK12-NEXT:  entry:
7314 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7315 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7316 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7317 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7318 // CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
7319 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7320 // CHECK12-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
7321 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7322 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7323 // CHECK12-NEXT:    ret void
7324 //
7325 //
7326 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
7327 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
7328 // CHECK12-NEXT:  entry:
7329 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7330 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7331 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7332 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7333 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7334 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7335 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7336 // CHECK12-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
7337 // CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
7338 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
7339 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
7340 // CHECK12-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
7341 // CHECK12-NEXT:    ret void
7342 //
7343 //
7344 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
7345 // CHECK12-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
7346 // CHECK12-NEXT:  entry:
7347 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7348 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7349 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7350 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7351 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7352 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7353 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7354 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
7355 // CHECK12-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
7356 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
7357 // CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
7358 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7359 // CHECK12-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
7360 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7361 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
7362 // CHECK12-NEXT:    ret void
7363 //
7364 //
7365 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
7366 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
7367 // CHECK12-NEXT:  entry:
7368 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7369 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7370 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7371 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7372 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7373 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7374 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7375 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7376 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7377 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
7378 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
7379 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7380 // CHECK12-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
7381 // CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
7382 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
7383 // CHECK12-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
7384 // CHECK12-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
7385 // CHECK12-NEXT:    ret void
7386 //
7387 //
7388 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
7389 // 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:%.*]]) #[[ATTR0]] {
7390 // CHECK12-NEXT:  entry:
7391 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7392 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
7393 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7394 // CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
7395 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
7396 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7397 // CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
7398 // CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
7399 // CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
7400 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7401 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7402 // CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
7403 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7404 // CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
7405 // CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
7406 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7407 // CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
7408 // CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
7409 // CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
7410 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
7411 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7412 // CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
7413 // CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
7414 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7415 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
7416 // CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
7417 // CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
7418 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
7419 // CHECK12-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
7420 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
7421 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
7422 // CHECK12-NEXT:    ret void
7423 //
7424 //
7425 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
7426 // 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:%.*]]) #[[ATTR0]] {
7427 // CHECK12-NEXT:  entry:
7428 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7429 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7430 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7431 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
7432 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7433 // CHECK12-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
7434 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
7435 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7436 // CHECK12-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
7437 // CHECK12-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
7438 // CHECK12-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
7439 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7440 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7441 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7442 // CHECK12-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
7443 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7444 // CHECK12-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
7445 // CHECK12-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
7446 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7447 // CHECK12-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
7448 // CHECK12-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
7449 // CHECK12-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
7450 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
7451 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7452 // CHECK12-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
7453 // CHECK12-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
7454 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7455 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
7456 // CHECK12-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
7457 // CHECK12-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
7458 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
7459 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
7460 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7461 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
7462 // CHECK12-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
7463 // CHECK12-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
7464 // CHECK12-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
7465 // CHECK12-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
7466 // CHECK12-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
7467 // CHECK12-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
7468 // CHECK12-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
7469 // CHECK12-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
7470 // CHECK12-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
7471 // CHECK12-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
7472 // CHECK12-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
7473 // CHECK12-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
7474 // CHECK12-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
7475 // CHECK12-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
7476 // CHECK12-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
7477 // CHECK12-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
7478 // CHECK12-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
7479 // CHECK12-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
7480 // CHECK12-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
7481 // CHECK12-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
7482 // CHECK12-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
7483 // CHECK12-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
7484 // CHECK12-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
7485 // CHECK12-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
7486 // CHECK12-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
7487 // CHECK12-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
7488 // CHECK12-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
7489 // CHECK12-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
7490 // CHECK12-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
7491 // CHECK12-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
7492 // CHECK12-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
7493 // CHECK12-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
7494 // CHECK12-NEXT:    ret void
7495 //
7496 //
7497 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
7498 // CHECK12-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7499 // CHECK12-NEXT:  entry:
7500 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7501 // CHECK12-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
7502 // CHECK12-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7503 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
7504 // CHECK12-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
7505 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
7506 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7507 // CHECK12-NEXT:    ret void
7508 //
7509 //
7510 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
7511 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7512 // CHECK12-NEXT:  entry:
7513 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7514 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7515 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7516 // CHECK12-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
7517 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7518 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7519 // CHECK12-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7520 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
7521 // CHECK12-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
7522 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
7523 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7524 // CHECK12-NEXT:    ret void
7525 //
7526 //
7527 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
7528 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7529 // CHECK12-NEXT:  entry:
7530 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7531 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7532 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7533 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7534 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7535 // CHECK12-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7536 // CHECK12-NEXT:    ret void
7537 //
7538 //
7539 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
7540 // CHECK12-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7541 // CHECK12-NEXT:  entry:
7542 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7543 // CHECK12-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
7544 // CHECK12-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7545 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
7546 // CHECK12-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
7547 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
7548 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]])
7549 // CHECK12-NEXT:    ret void
7550 //
7551 //
7552 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
7553 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
7554 // CHECK12-NEXT:  entry:
7555 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7556 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7557 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
7558 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7559 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7560 // CHECK12-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
7561 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
7562 // CHECK12-NEXT:    ret void
7563 //
7564 //
7565 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..7
7566 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
7567 // CHECK12-NEXT:  entry:
7568 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7569 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7570 // CHECK12-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
7571 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7572 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7573 // CHECK12-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
7574 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
7575 // CHECK12-NEXT:    ret void
7576 //
7577 //
7578 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
7579 // CHECK12-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
7580 // CHECK12-NEXT:  entry:
7581 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7582 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7583 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7584 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32 [[TMP0]])
7585 // CHECK12-NEXT:    ret void
7586 //
7587 //
7588 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..8
7589 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
7590 // CHECK12-NEXT:  entry:
7591 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7592 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7593 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7594 // CHECK12-NEXT:    [[F:%.*]] = alloca i32*, align 4
7595 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7596 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7597 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7598 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7599 // CHECK12-NEXT:    ret void
7600 //
7601 //
7602 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
7603 // CHECK12-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7604 // CHECK12-NEXT:  entry:
7605 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7606 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7607 // CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
7608 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7609 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7610 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7611 // CHECK12-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
7612 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7613 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7614 // CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
7615 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7616 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7617 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
7618 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7619 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7620 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
7621 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
7622 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7623 // CHECK12-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7624 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
7625 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7626 // CHECK12-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
7627 // CHECK12-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
7628 // CHECK12-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
7629 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
7630 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
7631 // CHECK12-NEXT:    ret void
7632 //
7633 //
7634 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..9
7635 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7636 // CHECK12-NEXT:  entry:
7637 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7638 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7639 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7640 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7641 // CHECK12-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
7642 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7643 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7644 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7645 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7646 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7647 // CHECK12-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
7648 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7649 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7650 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
7651 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7652 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7653 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
7654 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7655 // CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
7656 // CHECK12-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
7657 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
7658 // CHECK12-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
7659 // CHECK12-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
7660 // CHECK12-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
7661 // CHECK12-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
7662 // CHECK12-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
7663 // CHECK12-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
7664 // CHECK12-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
7665 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
7666 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7667 // CHECK12-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
7668 // CHECK12-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
7669 // CHECK12-NEXT:    ret void
7670 //
7671 //
7672 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
7673 // CHECK12-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
7674 // CHECK12-NEXT:  entry:
7675 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
7676 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
7677 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7678 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7679 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
7680 // CHECK12-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
7681 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
7682 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
7683 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7684 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7685 // CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
7686 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
7687 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7688 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7689 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
7690 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
7691 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
7692 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
7693 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
7694 // CHECK12-NEXT:    ret void
7695 //
7696 //
7697 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..10
7698 // 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]] {
7699 // CHECK12-NEXT:  entry:
7700 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7701 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7702 // CHECK12-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
7703 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
7704 // CHECK12-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
7705 // CHECK12-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
7706 // CHECK12-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
7707 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7708 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7709 // CHECK12-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
7710 // CHECK12-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
7711 // CHECK12-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7712 // CHECK12-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
7713 // CHECK12-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
7714 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
7715 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7716 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
7717 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
7718 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
7719 // CHECK12-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
7720 // CHECK12-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
7721 // CHECK12-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
7722 // CHECK12-NEXT:    store double [[ADD]], double* [[A]], align 4
7723 // CHECK12-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
7724 // CHECK12-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
7725 // CHECK12-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
7726 // CHECK12-NEXT:    store double [[INC]], double* [[A3]], align 4
7727 // CHECK12-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
7728 // CHECK12-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
7729 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
7730 // CHECK12-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
7731 // CHECK12-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
7732 // CHECK12-NEXT:    ret void
7733 //
7734 //
7735 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
7736 // CHECK12-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
7737 // CHECK12-NEXT:  entry:
7738 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7739 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7740 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7741 // CHECK12-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
7742 // CHECK12-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
7743 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7744 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7745 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7746 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7747 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7748 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7749 // CHECK12-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
7750 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
7751 // CHECK12-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
7752 // CHECK12-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
7753 // CHECK12-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
7754 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
7755 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
7756 // CHECK12-NEXT:    ret void
7757 //
7758 //
7759 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..11
7760 // 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]] {
7761 // CHECK12-NEXT:  entry:
7762 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7763 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7764 // CHECK12-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
7765 // CHECK12-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
7766 // CHECK12-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
7767 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7768 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7769 // CHECK12-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
7770 // CHECK12-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
7771 // CHECK12-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
7772 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
7773 // CHECK12-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
7774 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
7775 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
7776 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
7777 // CHECK12-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
7778 // CHECK12-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
7779 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
7780 // CHECK12-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
7781 // CHECK12-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
7782 // CHECK12-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
7783 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
7784 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
7785 // CHECK12-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
7786 // CHECK12-NEXT:    ret void
7787 //
7788 //
7789 // CHECK17-LABEL: define {{[^@]+}}@_Z3fooi
7790 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
7791 // CHECK17-NEXT:  entry:
7792 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7793 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
7794 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
7795 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
7796 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7797 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7798 // CHECK17-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
7799 // CHECK17-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
7800 // CHECK17-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
7801 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7802 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
7803 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
7804 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
7805 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
7806 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
7807 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
7808 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
7809 // CHECK17-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
7810 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
7811 // CHECK17-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
7812 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
7813 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
7814 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
7815 // CHECK17-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
7816 // CHECK17-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
7817 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
7818 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
7819 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
7820 // CHECK17-NEXT:    [[A_CASTED21:%.*]] = alloca i64, align 8
7821 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS25:%.*]] = alloca [9 x i8*], align 8
7822 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS26:%.*]] = alloca [9 x i8*], align 8
7823 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS27:%.*]] = alloca [9 x i8*], align 8
7824 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
7825 // CHECK17-NEXT:    [[NN:%.*]] = alloca i32, align 4
7826 // CHECK17-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
7827 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 8
7828 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 8
7829 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 8
7830 // CHECK17-NEXT:    [[NN_CASTED38:%.*]] = alloca i64, align 8
7831 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS40:%.*]] = alloca [1 x i8*], align 8
7832 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS41:%.*]] = alloca [1 x i8*], align 8
7833 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS42:%.*]] = alloca [1 x i8*], align 8
7834 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
7835 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7836 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
7837 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
7838 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7839 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
7840 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
7841 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
7842 // CHECK17-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
7843 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
7844 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7845 // CHECK17-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
7846 // CHECK17-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
7847 // CHECK17-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
7848 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
7849 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
7850 // CHECK17-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
7851 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7852 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
7853 // CHECK17-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
7854 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
7855 // CHECK17-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
7856 // CHECK17-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
7857 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7858 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
7859 // CHECK17-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
7860 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
7861 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7862 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
7863 // CHECK17-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
7864 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
7865 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7866 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
7867 // CHECK17-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
7868 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7869 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
7870 // CHECK17-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
7871 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7872 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
7873 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
7874 // CHECK17-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
7875 // CHECK17-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
7876 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
7877 // CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
7878 // CHECK17-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
7879 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
7880 // CHECK17-NEXT:    store i8* null, i8** [[TMP24]], align 8
7881 // CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
7882 // CHECK17-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
7883 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
7884 // CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
7885 // CHECK17-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
7886 // CHECK17-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
7887 // CHECK17-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
7888 // CHECK17-NEXT:    store i8* null, i8** [[TMP29]], align 8
7889 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7890 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7891 // CHECK17-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
7892 // CHECK17-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
7893 // CHECK17-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
7894 // CHECK17-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
7895 // CHECK17-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7896 // CHECK17-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
7897 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
7898 // CHECK17-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7899 // CHECK17-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
7900 // CHECK17-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
7901 // CHECK17-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
7902 // CHECK17-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
7903 // CHECK17-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
7904 // CHECK17-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
7905 // CHECK17-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
7906 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
7907 // CHECK17-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
7908 // CHECK17-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
7909 // CHECK17-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
7910 // CHECK17-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
7911 // CHECK17-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
7912 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
7913 // CHECK17-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
7914 // CHECK17-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
7915 // CHECK17-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
7916 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
7917 // CHECK17-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
7918 // CHECK17-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
7919 // 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)
7920 // CHECK17-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
7921 // CHECK17-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
7922 // CHECK17-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
7923 // CHECK17-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP38]])
7924 // CHECK17-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
7925 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
7926 // CHECK17-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
7927 // CHECK17-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
7928 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
7929 // CHECK17-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
7930 // CHECK17-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
7931 // CHECK17-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
7932 // CHECK17-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
7933 // CHECK17-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
7934 // CHECK17-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
7935 // CHECK17-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
7936 // CHECK17-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
7937 // CHECK17-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
7938 // CHECK17-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
7939 // CHECK17-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
7940 // CHECK17-NEXT:    store i8* null, i8** [[TMP65]], align 8
7941 // CHECK17-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
7942 // CHECK17-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
7943 // CHECK17-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
7944 // CHECK17-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
7945 // CHECK17-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7946 // CHECK17:       omp_offload.failed:
7947 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i64 [[TMP60]]) #[[ATTR3]]
7948 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7949 // CHECK17:       omp_offload.cont:
7950 // CHECK17-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
7951 // CHECK17-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
7952 // CHECK17-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
7953 // CHECK17-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
7954 // CHECK17-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
7955 // CHECK17-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
7956 // CHECK17-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
7957 // CHECK17-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
7958 // CHECK17-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
7959 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
7960 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7961 // CHECK17:       omp_if.then:
7962 // CHECK17-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
7963 // CHECK17-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
7964 // CHECK17-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
7965 // CHECK17-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
7966 // CHECK17-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
7967 // CHECK17-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
7968 // CHECK17-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
7969 // CHECK17-NEXT:    store i8* null, i8** [[TMP79]], align 8
7970 // CHECK17-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
7971 // CHECK17-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
7972 // CHECK17-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
7973 // CHECK17-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
7974 // CHECK17-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
7975 // CHECK17-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
7976 // CHECK17-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
7977 // CHECK17-NEXT:    store i8* null, i8** [[TMP84]], align 8
7978 // CHECK17-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
7979 // CHECK17-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
7980 // CHECK17-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
7981 // CHECK17-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
7982 // CHECK17-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
7983 // CHECK17:       omp_offload.failed19:
7984 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
7985 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
7986 // CHECK17:       omp_offload.cont20:
7987 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
7988 // CHECK17:       omp_if.else:
7989 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
7990 // CHECK17-NEXT:    br label [[OMP_IF_END]]
7991 // CHECK17:       omp_if.end:
7992 // CHECK17-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A]], align 4
7993 // CHECK17-NEXT:    [[CONV22:%.*]] = bitcast i64* [[A_CASTED21]] to i32*
7994 // CHECK17-NEXT:    store i32 [[TMP89]], i32* [[CONV22]], align 4
7995 // CHECK17-NEXT:    [[TMP90:%.*]] = load i64, i64* [[A_CASTED21]], align 8
7996 // CHECK17-NEXT:    [[TMP91:%.*]] = load i32, i32* [[N_ADDR]], align 4
7997 // CHECK17-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP91]], 20
7998 // CHECK17-NEXT:    br i1 [[CMP23]], label [[OMP_IF_THEN24:%.*]], label [[OMP_IF_ELSE30:%.*]]
7999 // CHECK17:       omp_if.then24:
8000 // CHECK17-NEXT:    [[TMP92:%.*]] = mul nuw i64 [[TMP2]], 4
8001 // CHECK17-NEXT:    [[TMP93:%.*]] = mul nuw i64 5, [[TMP5]]
8002 // CHECK17-NEXT:    [[TMP94:%.*]] = mul nuw i64 [[TMP93]], 8
8003 // CHECK17-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
8004 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP95]], i8* align 8 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i64 72, i1 false)
8005 // CHECK17-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
8006 // CHECK17-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i64*
8007 // CHECK17-NEXT:    store i64 [[TMP90]], i64* [[TMP97]], align 8
8008 // CHECK17-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
8009 // CHECK17-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i64*
8010 // CHECK17-NEXT:    store i64 [[TMP90]], i64* [[TMP99]], align 8
8011 // CHECK17-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 0
8012 // CHECK17-NEXT:    store i8* null, i8** [[TMP100]], align 8
8013 // CHECK17-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 1
8014 // CHECK17-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
8015 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 8
8016 // CHECK17-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 1
8017 // CHECK17-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
8018 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 8
8019 // CHECK17-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 1
8020 // CHECK17-NEXT:    store i8* null, i8** [[TMP105]], align 8
8021 // CHECK17-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 2
8022 // CHECK17-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
8023 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP107]], align 8
8024 // CHECK17-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 2
8025 // CHECK17-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i64*
8026 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP109]], align 8
8027 // CHECK17-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 2
8028 // CHECK17-NEXT:    store i8* null, i8** [[TMP110]], align 8
8029 // CHECK17-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 3
8030 // CHECK17-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
8031 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP112]], align 8
8032 // CHECK17-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 3
8033 // CHECK17-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
8034 // CHECK17-NEXT:    store float* [[VLA]], float** [[TMP114]], align 8
8035 // CHECK17-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
8036 // CHECK17-NEXT:    store i64 [[TMP92]], i64* [[TMP115]], align 8
8037 // CHECK17-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 3
8038 // CHECK17-NEXT:    store i8* null, i8** [[TMP116]], align 8
8039 // CHECK17-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 4
8040 // CHECK17-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
8041 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 8
8042 // CHECK17-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 4
8043 // CHECK17-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
8044 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 8
8045 // CHECK17-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 4
8046 // CHECK17-NEXT:    store i8* null, i8** [[TMP121]], align 8
8047 // CHECK17-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 5
8048 // CHECK17-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i64*
8049 // CHECK17-NEXT:    store i64 5, i64* [[TMP123]], align 8
8050 // CHECK17-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 5
8051 // CHECK17-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i64*
8052 // CHECK17-NEXT:    store i64 5, i64* [[TMP125]], align 8
8053 // CHECK17-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 5
8054 // CHECK17-NEXT:    store i8* null, i8** [[TMP126]], align 8
8055 // CHECK17-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 6
8056 // CHECK17-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
8057 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP128]], align 8
8058 // CHECK17-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 6
8059 // CHECK17-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i64*
8060 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP130]], align 8
8061 // CHECK17-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 6
8062 // CHECK17-NEXT:    store i8* null, i8** [[TMP131]], align 8
8063 // CHECK17-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 7
8064 // CHECK17-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
8065 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 8
8066 // CHECK17-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 7
8067 // CHECK17-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
8068 // CHECK17-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 8
8069 // CHECK17-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
8070 // CHECK17-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 8
8071 // CHECK17-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 7
8072 // CHECK17-NEXT:    store i8* null, i8** [[TMP137]], align 8
8073 // CHECK17-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 8
8074 // CHECK17-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
8075 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 8
8076 // CHECK17-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 8
8077 // CHECK17-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
8078 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 8
8079 // CHECK17-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 8
8080 // CHECK17-NEXT:    store i8* null, i8** [[TMP142]], align 8
8081 // CHECK17-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
8082 // CHECK17-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
8083 // CHECK17-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
8084 // CHECK17-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8085 // CHECK17-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
8086 // CHECK17-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
8087 // CHECK17:       omp_offload.failed28:
8088 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
8089 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
8090 // CHECK17:       omp_offload.cont29:
8091 // CHECK17-NEXT:    br label [[OMP_IF_END31:%.*]]
8092 // CHECK17:       omp_if.else30:
8093 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
8094 // CHECK17-NEXT:    br label [[OMP_IF_END31]]
8095 // CHECK17:       omp_if.end31:
8096 // CHECK17-NEXT:    store i32 0, i32* [[NN]], align 4
8097 // CHECK17-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
8098 // CHECK17-NEXT:    [[CONV32:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
8099 // CHECK17-NEXT:    store i32 [[TMP148]], i32* [[CONV32]], align 4
8100 // CHECK17-NEXT:    [[TMP149:%.*]] = load i64, i64* [[NN_CASTED]], align 8
8101 // CHECK17-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
8102 // CHECK17-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
8103 // CHECK17-NEXT:    store i64 [[TMP149]], i64* [[TMP151]], align 8
8104 // CHECK17-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
8105 // CHECK17-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i64*
8106 // CHECK17-NEXT:    store i64 [[TMP149]], i64* [[TMP153]], align 8
8107 // CHECK17-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i64 0, i64 0
8108 // CHECK17-NEXT:    store i8* null, i8** [[TMP154]], align 8
8109 // CHECK17-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
8110 // CHECK17-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
8111 // CHECK17-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8112 // CHECK17-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
8113 // CHECK17-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
8114 // CHECK17:       omp_offload.failed36:
8115 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i64 [[TMP149]]) #[[ATTR3]]
8116 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
8117 // CHECK17:       omp_offload.cont37:
8118 // CHECK17-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
8119 // CHECK17-NEXT:    [[CONV39:%.*]] = bitcast i64* [[NN_CASTED38]] to i32*
8120 // CHECK17-NEXT:    store i32 [[TMP159]], i32* [[CONV39]], align 4
8121 // CHECK17-NEXT:    [[TMP160:%.*]] = load i64, i64* [[NN_CASTED38]], align 8
8122 // CHECK17-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
8123 // CHECK17-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i64*
8124 // CHECK17-NEXT:    store i64 [[TMP160]], i64* [[TMP162]], align 8
8125 // CHECK17-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
8126 // CHECK17-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i64*
8127 // CHECK17-NEXT:    store i64 [[TMP160]], i64* [[TMP164]], align 8
8128 // CHECK17-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS42]], i64 0, i64 0
8129 // CHECK17-NEXT:    store i8* null, i8** [[TMP165]], align 8
8130 // CHECK17-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
8131 // CHECK17-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
8132 // CHECK17-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8133 // CHECK17-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
8134 // CHECK17-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
8135 // CHECK17:       omp_offload.failed43:
8136 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i64 [[TMP160]]) #[[ATTR3]]
8137 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
8138 // CHECK17:       omp_offload.cont44:
8139 // CHECK17-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
8140 // CHECK17-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
8141 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
8142 // CHECK17-NEXT:    ret i32 [[TMP170]]
8143 //
8144 //
8145 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
8146 // CHECK17-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
8147 // CHECK17-NEXT:  entry:
8148 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8149 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8150 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
8151 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8152 // CHECK17-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
8153 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8154 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8155 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
8156 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8157 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8158 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
8159 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
8160 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
8161 // CHECK17-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
8162 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
8163 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8164 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
8165 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8166 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
8167 // CHECK17-NEXT:    ret void
8168 //
8169 //
8170 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
8171 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
8172 // CHECK17-NEXT:  entry:
8173 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8174 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8175 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8176 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8177 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8178 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8179 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8180 // CHECK17-NEXT:    ret void
8181 //
8182 //
8183 // CHECK17-LABEL: define {{[^@]+}}@.omp_task_privates_map.
8184 // 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]+]] {
8185 // CHECK17-NEXT:  entry:
8186 // CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
8187 // CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
8188 // CHECK17-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
8189 // CHECK17-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
8190 // CHECK17-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
8191 // CHECK17-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
8192 // CHECK17-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
8193 // CHECK17-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
8194 // CHECK17-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
8195 // CHECK17-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
8196 // CHECK17-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
8197 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
8198 // CHECK17-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
8199 // CHECK17-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
8200 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
8201 // CHECK17-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
8202 // CHECK17-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
8203 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
8204 // CHECK17-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
8205 // CHECK17-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
8206 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
8207 // CHECK17-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
8208 // CHECK17-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
8209 // CHECK17-NEXT:    ret void
8210 //
8211 //
8212 // CHECK17-LABEL: define {{[^@]+}}@.omp_task_entry.
8213 // CHECK17-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
8214 // CHECK17-NEXT:  entry:
8215 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
8216 // CHECK17-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
8217 // CHECK17-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
8218 // CHECK17-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
8219 // CHECK17-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
8220 // CHECK17-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
8221 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
8222 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
8223 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
8224 // CHECK17-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
8225 // CHECK17-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
8226 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
8227 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
8228 // CHECK17-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
8229 // CHECK17-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
8230 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
8231 // CHECK17-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
8232 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
8233 // CHECK17-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
8234 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
8235 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
8236 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
8237 // CHECK17-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
8238 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
8239 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
8240 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
8241 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
8242 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
8243 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
8244 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
8245 // CHECK17-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
8246 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !24
8247 // CHECK17-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !24
8248 // CHECK17-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
8249 // 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 !24
8250 // CHECK17-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !24
8251 // CHECK17-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
8252 // CHECK17-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
8253 // CHECK17-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !24
8254 // CHECK17-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
8255 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
8256 // 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]]
8257 // CHECK17-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !24
8258 // CHECK17-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !24
8259 // CHECK17-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !24
8260 // CHECK17-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !24
8261 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
8262 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
8263 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
8264 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
8265 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
8266 // CHECK17-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
8267 // CHECK17-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
8268 // CHECK17-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
8269 // CHECK17-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
8270 // CHECK17-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
8271 // CHECK17:       omp_offload.failed.i:
8272 // CHECK17-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
8273 // CHECK17-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
8274 // CHECK17-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !24
8275 // CHECK17-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !24
8276 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
8277 // CHECK17-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
8278 // CHECK17-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !24
8279 // CHECK17-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !24
8280 // CHECK17-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
8281 // CHECK17-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
8282 // CHECK17-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !24
8283 // CHECK17-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !24
8284 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
8285 // CHECK17-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
8286 // CHECK17:       .omp_outlined..1.exit:
8287 // CHECK17-NEXT:    ret i32 0
8288 //
8289 //
8290 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
8291 // CHECK17-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
8292 // CHECK17-NEXT:  entry:
8293 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8294 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8295 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8296 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8297 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8298 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8299 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
8300 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8301 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
8302 // CHECK17-NEXT:    ret void
8303 //
8304 //
8305 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
8306 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
8307 // CHECK17-NEXT:  entry:
8308 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8309 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8310 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8311 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8312 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8313 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8314 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8315 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8316 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
8317 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
8318 // CHECK17-NEXT:    ret void
8319 //
8320 //
8321 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
8322 // CHECK17-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
8323 // CHECK17-NEXT:  entry:
8324 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8325 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8326 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8327 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8328 // CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
8329 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8330 // CHECK17-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
8331 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8332 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
8333 // CHECK17-NEXT:    ret void
8334 //
8335 //
8336 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
8337 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
8338 // CHECK17-NEXT:  entry:
8339 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8340 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8341 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8342 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8343 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8344 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8345 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8346 // CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
8347 // CHECK17-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
8348 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
8349 // CHECK17-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
8350 // CHECK17-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
8351 // CHECK17-NEXT:    ret void
8352 //
8353 //
8354 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
8355 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
8356 // CHECK17-NEXT:  entry:
8357 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8358 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8359 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8360 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8361 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8362 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8363 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8364 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8365 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8366 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8367 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
8368 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8369 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
8370 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8371 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
8372 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8373 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
8374 // CHECK17-NEXT:    ret void
8375 //
8376 //
8377 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
8378 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
8379 // CHECK17-NEXT:  entry:
8380 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8381 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8382 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8383 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8384 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8385 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8386 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8387 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8388 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8389 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8390 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8391 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
8392 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
8393 // CHECK17-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
8394 // CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
8395 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
8396 // CHECK17-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
8397 // CHECK17-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
8398 // CHECK17-NEXT:    ret void
8399 //
8400 //
8401 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
8402 // 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:%.*]]) #[[ATTR2]] {
8403 // CHECK17-NEXT:  entry:
8404 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8405 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
8406 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8407 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
8408 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
8409 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8410 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
8411 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
8412 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
8413 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8414 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8415 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
8416 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8417 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
8418 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
8419 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8420 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
8421 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
8422 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
8423 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8424 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
8425 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8426 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
8427 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
8428 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8429 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
8430 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
8431 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
8432 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
8433 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8434 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
8435 // CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
8436 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
8437 // CHECK17-NEXT:    ret void
8438 //
8439 //
8440 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
8441 // 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:%.*]]) #[[ATTR2]] {
8442 // CHECK17-NEXT:  entry:
8443 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8444 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8445 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8446 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
8447 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8448 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
8449 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
8450 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8451 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
8452 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
8453 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
8454 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8455 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8456 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8457 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
8458 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8459 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
8460 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
8461 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8462 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
8463 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
8464 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
8465 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8466 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
8467 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8468 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
8469 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
8470 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8471 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
8472 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
8473 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
8474 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
8475 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
8476 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
8477 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
8478 // CHECK17-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
8479 // CHECK17-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
8480 // CHECK17-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
8481 // CHECK17-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
8482 // CHECK17-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
8483 // CHECK17-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
8484 // CHECK17-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
8485 // CHECK17-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
8486 // CHECK17-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
8487 // CHECK17-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
8488 // CHECK17-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
8489 // CHECK17-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
8490 // CHECK17-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
8491 // CHECK17-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
8492 // CHECK17-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
8493 // CHECK17-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
8494 // CHECK17-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
8495 // CHECK17-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
8496 // CHECK17-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
8497 // CHECK17-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
8498 // CHECK17-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
8499 // CHECK17-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
8500 // CHECK17-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
8501 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
8502 // CHECK17-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
8503 // CHECK17-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
8504 // CHECK17-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
8505 // CHECK17-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
8506 // CHECK17-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
8507 // CHECK17-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
8508 // CHECK17-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
8509 // CHECK17-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
8510 // CHECK17-NEXT:    ret void
8511 //
8512 //
8513 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
8514 // CHECK17-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
8515 // CHECK17-NEXT:  entry:
8516 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
8517 // CHECK17-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
8518 // CHECK17-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
8519 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
8520 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8521 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
8522 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
8523 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
8524 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP1]])
8525 // CHECK17-NEXT:    ret void
8526 //
8527 //
8528 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..12
8529 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
8530 // CHECK17-NEXT:  entry:
8531 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8532 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8533 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
8534 // CHECK17-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
8535 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8536 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8537 // CHECK17-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
8538 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
8539 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8540 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
8541 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
8542 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
8543 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP1]])
8544 // CHECK17-NEXT:    ret void
8545 //
8546 //
8547 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..13
8548 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
8549 // CHECK17-NEXT:  entry:
8550 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8551 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8552 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
8553 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8554 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8555 // CHECK17-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
8556 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
8557 // CHECK17-NEXT:    ret void
8558 //
8559 //
8560 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
8561 // CHECK17-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
8562 // CHECK17-NEXT:  entry:
8563 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
8564 // CHECK17-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
8565 // CHECK17-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
8566 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
8567 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8568 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
8569 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
8570 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
8571 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP1]])
8572 // CHECK17-NEXT:    ret void
8573 //
8574 //
8575 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..16
8576 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
8577 // CHECK17-NEXT:  entry:
8578 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8579 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8580 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
8581 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8582 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8583 // CHECK17-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
8584 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
8585 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[CONV]])
8586 // CHECK17-NEXT:    ret void
8587 //
8588 //
8589 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..17
8590 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
8591 // CHECK17-NEXT:  entry:
8592 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8593 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8594 // CHECK17-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
8595 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8596 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8597 // CHECK17-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
8598 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
8599 // CHECK17-NEXT:    ret void
8600 //
8601 //
8602 // CHECK17-LABEL: define {{[^@]+}}@_Z6bazzzziPi
8603 // CHECK17-SAME: (i32 noundef signext [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
8604 // CHECK17-NEXT:  entry:
8605 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8606 // CHECK17-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 8
8607 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
8608 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
8609 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
8610 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8611 // CHECK17-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 8
8612 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8613 // CHECK17-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
8614 // CHECK17-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8615 // CHECK17-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
8616 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
8617 // CHECK17-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8618 // CHECK17-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
8619 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
8620 // CHECK17-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8621 // CHECK17-NEXT:    store i8* null, i8** [[TMP6]], align 8
8622 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8623 // CHECK17-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8624 // CHECK17-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8625 // CHECK17-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
8626 // CHECK17-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8627 // CHECK17:       omp_offload.failed:
8628 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i64 [[TMP1]]) #[[ATTR3]]
8629 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8630 // CHECK17:       omp_offload.cont:
8631 // CHECK17-NEXT:    ret void
8632 //
8633 //
8634 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
8635 // CHECK17-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
8636 // CHECK17-NEXT:  entry:
8637 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8638 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8639 // CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8640 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i64 [[TMP0]])
8641 // CHECK17-NEXT:    ret void
8642 //
8643 //
8644 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..20
8645 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
8646 // CHECK17-NEXT:  entry:
8647 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8648 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8649 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8650 // CHECK17-NEXT:    [[F:%.*]] = alloca i32*, align 8
8651 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8652 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8653 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8654 // CHECK17-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8655 // CHECK17-NEXT:    ret void
8656 //
8657 //
8658 // CHECK17-LABEL: define {{[^@]+}}@_Z3bari
8659 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
8660 // CHECK17-NEXT:  entry:
8661 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8662 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
8663 // CHECK17-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
8664 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8665 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
8666 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8667 // CHECK17-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
8668 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
8669 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
8670 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8671 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
8672 // CHECK17-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
8673 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
8674 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
8675 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
8676 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8677 // CHECK17-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
8678 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
8679 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
8680 // CHECK17-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
8681 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
8682 // CHECK17-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
8683 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
8684 // CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
8685 // CHECK17-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
8686 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
8687 // CHECK17-NEXT:    ret i32 [[TMP8]]
8688 //
8689 //
8690 // CHECK17-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
8691 // CHECK17-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
8692 // CHECK17-NEXT:  entry:
8693 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8694 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8695 // CHECK17-NEXT:    [[B:%.*]] = alloca i32, align 4
8696 // CHECK17-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
8697 // CHECK17-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
8698 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
8699 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
8700 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
8701 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
8702 // CHECK17-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
8703 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8704 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8705 // CHECK17-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8706 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8707 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
8708 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
8709 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
8710 // CHECK17-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
8711 // CHECK17-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
8712 // CHECK17-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
8713 // CHECK17-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
8714 // CHECK17-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
8715 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
8716 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
8717 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
8718 // CHECK17-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
8719 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
8720 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
8721 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
8722 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8723 // CHECK17:       omp_if.then:
8724 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
8725 // CHECK17-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
8726 // CHECK17-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
8727 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
8728 // CHECK17-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i64 40, i1 false)
8729 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8730 // CHECK17-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
8731 // CHECK17-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
8732 // CHECK17-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8733 // CHECK17-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
8734 // CHECK17-NEXT:    store double* [[A]], double** [[TMP14]], align 8
8735 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8736 // CHECK17-NEXT:    store i8* null, i8** [[TMP15]], align 8
8737 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8738 // CHECK17-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
8739 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
8740 // CHECK17-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8741 // CHECK17-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
8742 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
8743 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8744 // CHECK17-NEXT:    store i8* null, i8** [[TMP20]], align 8
8745 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8746 // CHECK17-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
8747 // CHECK17-NEXT:    store i64 2, i64* [[TMP22]], align 8
8748 // CHECK17-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8749 // CHECK17-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
8750 // CHECK17-NEXT:    store i64 2, i64* [[TMP24]], align 8
8751 // CHECK17-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8752 // CHECK17-NEXT:    store i8* null, i8** [[TMP25]], align 8
8753 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
8754 // CHECK17-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
8755 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
8756 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
8757 // CHECK17-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
8758 // CHECK17-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
8759 // CHECK17-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
8760 // CHECK17-NEXT:    store i8* null, i8** [[TMP30]], align 8
8761 // CHECK17-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
8762 // CHECK17-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
8763 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
8764 // CHECK17-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
8765 // CHECK17-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
8766 // CHECK17-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
8767 // CHECK17-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
8768 // CHECK17-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
8769 // CHECK17-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
8770 // CHECK17-NEXT:    store i8* null, i8** [[TMP36]], align 8
8771 // CHECK17-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8772 // CHECK17-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8773 // CHECK17-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
8774 // CHECK17-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8775 // CHECK17-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
8776 // CHECK17-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8777 // CHECK17:       omp_offload.failed:
8778 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
8779 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8780 // CHECK17:       omp_offload.cont:
8781 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
8782 // CHECK17:       omp_if.else:
8783 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
8784 // CHECK17-NEXT:    br label [[OMP_IF_END]]
8785 // CHECK17:       omp_if.end:
8786 // CHECK17-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
8787 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
8788 // CHECK17-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
8789 // CHECK17-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
8790 // CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
8791 // CHECK17-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
8792 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
8793 // CHECK17-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
8794 // CHECK17-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
8795 // CHECK17-NEXT:    ret i32 [[ADD4]]
8796 //
8797 //
8798 // CHECK17-LABEL: define {{[^@]+}}@_ZL7fstatici
8799 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
8800 // CHECK17-NEXT:  entry:
8801 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8802 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
8803 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
8804 // CHECK17-NEXT:    [[AAA:%.*]] = alloca i8, align 1
8805 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8806 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8807 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8808 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
8809 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
8810 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
8811 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
8812 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8813 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
8814 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
8815 // CHECK17-NEXT:    store i8 0, i8* [[AAA]], align 1
8816 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8817 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8818 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
8819 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8820 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
8821 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8822 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
8823 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8824 // CHECK17-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
8825 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
8826 // CHECK17-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
8827 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
8828 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
8829 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
8830 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8831 // CHECK17:       omp_if.then:
8832 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8833 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
8834 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
8835 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8836 // CHECK17-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
8837 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
8838 // CHECK17-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8839 // CHECK17-NEXT:    store i8* null, i8** [[TMP11]], align 8
8840 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8841 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
8842 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
8843 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8844 // CHECK17-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
8845 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
8846 // CHECK17-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8847 // CHECK17-NEXT:    store i8* null, i8** [[TMP16]], align 8
8848 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8849 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
8850 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
8851 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8852 // CHECK17-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
8853 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
8854 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8855 // CHECK17-NEXT:    store i8* null, i8** [[TMP21]], align 8
8856 // CHECK17-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
8857 // CHECK17-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
8858 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
8859 // CHECK17-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
8860 // CHECK17-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
8861 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
8862 // CHECK17-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
8863 // CHECK17-NEXT:    store i8* null, i8** [[TMP26]], align 8
8864 // CHECK17-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8865 // CHECK17-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8866 // CHECK17-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8867 // CHECK17-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
8868 // CHECK17-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8869 // CHECK17:       omp_offload.failed:
8870 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
8871 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8872 // CHECK17:       omp_offload.cont:
8873 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
8874 // CHECK17:       omp_if.else:
8875 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
8876 // CHECK17-NEXT:    br label [[OMP_IF_END]]
8877 // CHECK17:       omp_if.end:
8878 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
8879 // CHECK17-NEXT:    ret i32 [[TMP31]]
8880 //
8881 //
8882 // CHECK17-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
8883 // CHECK17-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
8884 // CHECK17-NEXT:  entry:
8885 // CHECK17-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8886 // CHECK17-NEXT:    [[A:%.*]] = alloca i32, align 4
8887 // CHECK17-NEXT:    [[AA:%.*]] = alloca i16, align 2
8888 // CHECK17-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8889 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8890 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8891 // CHECK17-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
8892 // CHECK17-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
8893 // CHECK17-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
8894 // CHECK17-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8895 // CHECK17-NEXT:    store i32 0, i32* [[A]], align 4
8896 // CHECK17-NEXT:    store i16 0, i16* [[AA]], align 2
8897 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8898 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8899 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
8900 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8901 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
8902 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8903 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
8904 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8905 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8906 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
8907 // CHECK17-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8908 // CHECK17:       omp_if.then:
8909 // CHECK17-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8910 // CHECK17-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
8911 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
8912 // CHECK17-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8913 // CHECK17-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
8914 // CHECK17-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
8915 // CHECK17-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8916 // CHECK17-NEXT:    store i8* null, i8** [[TMP9]], align 8
8917 // CHECK17-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
8918 // CHECK17-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
8919 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
8920 // CHECK17-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
8921 // CHECK17-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
8922 // CHECK17-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
8923 // CHECK17-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
8924 // CHECK17-NEXT:    store i8* null, i8** [[TMP14]], align 8
8925 // CHECK17-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
8926 // CHECK17-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
8927 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
8928 // CHECK17-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
8929 // CHECK17-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
8930 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
8931 // CHECK17-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
8932 // CHECK17-NEXT:    store i8* null, i8** [[TMP19]], align 8
8933 // CHECK17-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8934 // CHECK17-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8935 // CHECK17-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8936 // CHECK17-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
8937 // CHECK17-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8938 // CHECK17:       omp_offload.failed:
8939 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
8940 // CHECK17-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8941 // CHECK17:       omp_offload.cont:
8942 // CHECK17-NEXT:    br label [[OMP_IF_END:%.*]]
8943 // CHECK17:       omp_if.else:
8944 // CHECK17-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
8945 // CHECK17-NEXT:    br label [[OMP_IF_END]]
8946 // CHECK17:       omp_if.end:
8947 // CHECK17-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
8948 // CHECK17-NEXT:    ret i32 [[TMP24]]
8949 //
8950 //
8951 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
8952 // CHECK17-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
8953 // CHECK17-NEXT:  entry:
8954 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8955 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
8956 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8957 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8958 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
8959 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
8960 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8961 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
8962 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8963 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8964 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
8965 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8966 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
8967 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8968 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8969 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
8970 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
8971 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
8972 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
8973 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
8974 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
8975 // CHECK17-NEXT:    ret void
8976 //
8977 //
8978 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..23
8979 // 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]] {
8980 // CHECK17-NEXT:  entry:
8981 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8982 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8983 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
8984 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
8985 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
8986 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
8987 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
8988 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8989 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8990 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
8991 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
8992 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8993 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
8994 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
8995 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
8996 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
8997 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8998 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
8999 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
9000 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
9001 // CHECK17-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
9002 // CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
9003 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
9004 // CHECK17-NEXT:    store double [[ADD]], double* [[A]], align 8
9005 // CHECK17-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
9006 // CHECK17-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
9007 // CHECK17-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
9008 // CHECK17-NEXT:    store double [[INC]], double* [[A4]], align 8
9009 // CHECK17-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
9010 // CHECK17-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
9011 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
9012 // CHECK17-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
9013 // CHECK17-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
9014 // CHECK17-NEXT:    ret void
9015 //
9016 //
9017 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
9018 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
9019 // CHECK17-NEXT:  entry:
9020 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9021 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9022 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
9023 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9024 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9025 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9026 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
9027 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9028 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9029 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
9030 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9031 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9032 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9033 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
9034 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9035 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9036 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9037 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
9038 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
9039 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
9040 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9041 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
9042 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9043 // CHECK17-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
9044 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
9045 // CHECK17-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
9046 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
9047 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
9048 // CHECK17-NEXT:    ret void
9049 //
9050 //
9051 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..26
9052 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
9053 // CHECK17-NEXT:  entry:
9054 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9055 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9056 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9057 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9058 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
9059 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9060 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9061 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9062 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9063 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9064 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
9065 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9066 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9067 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9068 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
9069 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9070 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9071 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
9072 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
9073 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
9074 // CHECK17-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
9075 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
9076 // CHECK17-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
9077 // CHECK17-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
9078 // CHECK17-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
9079 // CHECK17-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
9080 // CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
9081 // CHECK17-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
9082 // CHECK17-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
9083 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
9084 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9085 // CHECK17-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
9086 // CHECK17-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
9087 // CHECK17-NEXT:    ret void
9088 //
9089 //
9090 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
9091 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
9092 // CHECK17-NEXT:  entry:
9093 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9094 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9095 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9096 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9097 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9098 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9099 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9100 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9101 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9102 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9103 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9104 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9105 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9106 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
9107 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
9108 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
9109 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9110 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
9111 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9112 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
9113 // CHECK17-NEXT:    ret void
9114 //
9115 //
9116 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..29
9117 // 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]] {
9118 // CHECK17-NEXT:  entry:
9119 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9120 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9121 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9122 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9123 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9124 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9125 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9126 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9127 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9128 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9129 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9130 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9131 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9132 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9133 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
9134 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
9135 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
9136 // CHECK17-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
9137 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
9138 // CHECK17-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
9139 // CHECK17-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
9140 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
9141 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
9142 // CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
9143 // CHECK17-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
9144 // CHECK17-NEXT:    ret void
9145 //
9146 //
9147 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9148 // CHECK17-SAME: () #[[ATTR4]] {
9149 // CHECK17-NEXT:  entry:
9150 // CHECK17-NEXT:    call void @__tgt_register_requires(i64 1)
9151 // CHECK17-NEXT:    ret void
9152 //
9153 //
9154 // CHECK18-LABEL: define {{[^@]+}}@_Z3fooi
9155 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
9156 // CHECK18-NEXT:  entry:
9157 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9158 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
9159 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
9160 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
9161 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
9162 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
9163 // CHECK18-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
9164 // CHECK18-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
9165 // CHECK18-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
9166 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9167 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
9168 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9169 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9170 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED4:%.*]] = alloca i64, align 8
9171 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
9172 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
9173 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
9174 // CHECK18-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
9175 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9176 // CHECK18-NEXT:    [[AA_CASTED7:%.*]] = alloca i64, align 8
9177 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [1 x i8*], align 8
9178 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS10:%.*]] = alloca [1 x i8*], align 8
9179 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [1 x i8*], align 8
9180 // CHECK18-NEXT:    [[A_CASTED12:%.*]] = alloca i64, align 8
9181 // CHECK18-NEXT:    [[AA_CASTED14:%.*]] = alloca i64, align 8
9182 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS16:%.*]] = alloca [2 x i8*], align 8
9183 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS17:%.*]] = alloca [2 x i8*], align 8
9184 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS18:%.*]] = alloca [2 x i8*], align 8
9185 // CHECK18-NEXT:    [[A_CASTED21:%.*]] = alloca i64, align 8
9186 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS25:%.*]] = alloca [9 x i8*], align 8
9187 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS26:%.*]] = alloca [9 x i8*], align 8
9188 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS27:%.*]] = alloca [9 x i8*], align 8
9189 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 8
9190 // CHECK18-NEXT:    [[NN:%.*]] = alloca i32, align 4
9191 // CHECK18-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
9192 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 8
9193 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 8
9194 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 8
9195 // CHECK18-NEXT:    [[NN_CASTED38:%.*]] = alloca i64, align 8
9196 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS40:%.*]] = alloca [1 x i8*], align 8
9197 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS41:%.*]] = alloca [1 x i8*], align 8
9198 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS42:%.*]] = alloca [1 x i8*], align 8
9199 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
9200 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9201 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
9202 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
9203 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
9204 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
9205 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
9206 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
9207 // CHECK18-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
9208 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
9209 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
9210 // CHECK18-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
9211 // CHECK18-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
9212 // CHECK18-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
9213 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
9214 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
9215 // CHECK18-NEXT:    store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_]], align 4
9216 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
9217 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[DOTCAPTURE_EXPR_2]], align 4
9218 // CHECK18-NEXT:    [[TMP9:%.*]] = load i16, i16* [[AA]], align 2
9219 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9220 // CHECK18-NEXT:    store i16 [[TMP9]], i16* [[CONV]], align 2
9221 // CHECK18-NEXT:    [[TMP10:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9222 // CHECK18-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9223 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9224 // CHECK18-NEXT:    store i32 [[TMP11]], i32* [[CONV3]], align 4
9225 // CHECK18-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9226 // CHECK18-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
9227 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED4]] to i32*
9228 // CHECK18-NEXT:    store i32 [[TMP13]], i32* [[CONV5]], align 4
9229 // CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED4]], align 8
9230 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9231 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
9232 // CHECK18-NEXT:    store i64 [[TMP10]], i64* [[TMP16]], align 8
9233 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9234 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
9235 // CHECK18-NEXT:    store i64 [[TMP10]], i64* [[TMP18]], align 8
9236 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9237 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
9238 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
9239 // CHECK18-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
9240 // CHECK18-NEXT:    store i64 [[TMP12]], i64* [[TMP21]], align 8
9241 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
9242 // CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
9243 // CHECK18-NEXT:    store i64 [[TMP12]], i64* [[TMP23]], align 8
9244 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
9245 // CHECK18-NEXT:    store i8* null, i8** [[TMP24]], align 8
9246 // CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
9247 // CHECK18-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
9248 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP26]], align 8
9249 // CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
9250 // CHECK18-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i64*
9251 // CHECK18-NEXT:    store i64 [[TMP14]], i64* [[TMP28]], align 8
9252 // CHECK18-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
9253 // CHECK18-NEXT:    store i8* null, i8** [[TMP29]], align 8
9254 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9255 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9256 // CHECK18-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
9257 // CHECK18-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2
9258 // CHECK18-NEXT:    store i16 [[TMP33]], i16* [[TMP32]], align 4
9259 // CHECK18-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
9260 // CHECK18-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9261 // CHECK18-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
9262 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
9263 // CHECK18-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
9264 // CHECK18-NEXT:    store i32 [[TMP37]], i32* [[TMP36]], align 4
9265 // CHECK18-NEXT:    [[TMP38:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
9266 // CHECK18-NEXT:    [[TMP39:%.*]] = bitcast i8* [[TMP38]] to %struct.kmp_task_t_with_privates*
9267 // CHECK18-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 0
9268 // CHECK18-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP40]], i32 0, i32 0
9269 // CHECK18-NEXT:    [[TMP42:%.*]] = load i8*, i8** [[TMP41]], align 8
9270 // CHECK18-NEXT:    [[TMP43:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
9271 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 [[TMP42]], i8* align 4 [[TMP43]], i64 12, i1 false)
9272 // CHECK18-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP39]], i32 0, i32 1
9273 // CHECK18-NEXT:    [[TMP45:%.*]] = bitcast i8* [[TMP42]] to %struct.anon*
9274 // CHECK18-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 0
9275 // CHECK18-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
9276 // CHECK18-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP30]] to i8*
9277 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP47]], i8* align 8 [[TMP48]], i64 24, i1 false)
9278 // CHECK18-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 1
9279 // CHECK18-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
9280 // CHECK18-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP31]] to i8*
9281 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP50]], i8* align 8 [[TMP51]], i64 24, i1 false)
9282 // CHECK18-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 2
9283 // CHECK18-NEXT:    [[TMP53:%.*]] = bitcast [3 x i64]* [[TMP52]] to i8*
9284 // 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)
9285 // CHECK18-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP44]], i32 0, i32 3
9286 // CHECK18-NEXT:    [[TMP55:%.*]] = load i16, i16* [[AA]], align 2
9287 // CHECK18-NEXT:    store i16 [[TMP55]], i16* [[TMP54]], align 8
9288 // CHECK18-NEXT:    [[TMP56:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP38]])
9289 // CHECK18-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A]], align 4
9290 // CHECK18-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9291 // CHECK18-NEXT:    store i32 [[TMP57]], i32* [[CONV6]], align 4
9292 // CHECK18-NEXT:    [[TMP58:%.*]] = load i64, i64* [[A_CASTED]], align 8
9293 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i64 [[TMP58]]) #[[ATTR3:[0-9]+]]
9294 // CHECK18-NEXT:    [[TMP59:%.*]] = load i16, i16* [[AA]], align 2
9295 // CHECK18-NEXT:    [[CONV8:%.*]] = bitcast i64* [[AA_CASTED7]] to i16*
9296 // CHECK18-NEXT:    store i16 [[TMP59]], i16* [[CONV8]], align 2
9297 // CHECK18-NEXT:    [[TMP60:%.*]] = load i64, i64* [[AA_CASTED7]], align 8
9298 // CHECK18-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
9299 // CHECK18-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i64*
9300 // CHECK18-NEXT:    store i64 [[TMP60]], i64* [[TMP62]], align 8
9301 // CHECK18-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
9302 // CHECK18-NEXT:    [[TMP64:%.*]] = bitcast i8** [[TMP63]] to i64*
9303 // CHECK18-NEXT:    store i64 [[TMP60]], i64* [[TMP64]], align 8
9304 // CHECK18-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
9305 // CHECK18-NEXT:    store i8* null, i8** [[TMP65]], align 8
9306 // CHECK18-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
9307 // CHECK18-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
9308 // CHECK18-NEXT:    [[TMP68:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
9309 // CHECK18-NEXT:    [[TMP69:%.*]] = icmp ne i32 [[TMP68]], 0
9310 // CHECK18-NEXT:    br i1 [[TMP69]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9311 // CHECK18:       omp_offload.failed:
9312 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i64 [[TMP60]]) #[[ATTR3]]
9313 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9314 // CHECK18:       omp_offload.cont:
9315 // CHECK18-NEXT:    [[TMP70:%.*]] = load i32, i32* [[A]], align 4
9316 // CHECK18-NEXT:    [[CONV13:%.*]] = bitcast i64* [[A_CASTED12]] to i32*
9317 // CHECK18-NEXT:    store i32 [[TMP70]], i32* [[CONV13]], align 4
9318 // CHECK18-NEXT:    [[TMP71:%.*]] = load i64, i64* [[A_CASTED12]], align 8
9319 // CHECK18-NEXT:    [[TMP72:%.*]] = load i16, i16* [[AA]], align 2
9320 // CHECK18-NEXT:    [[CONV15:%.*]] = bitcast i64* [[AA_CASTED14]] to i16*
9321 // CHECK18-NEXT:    store i16 [[TMP72]], i16* [[CONV15]], align 2
9322 // CHECK18-NEXT:    [[TMP73:%.*]] = load i64, i64* [[AA_CASTED14]], align 8
9323 // CHECK18-NEXT:    [[TMP74:%.*]] = load i32, i32* [[N_ADDR]], align 4
9324 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP74]], 10
9325 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9326 // CHECK18:       omp_if.then:
9327 // CHECK18-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
9328 // CHECK18-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i64*
9329 // CHECK18-NEXT:    store i64 [[TMP71]], i64* [[TMP76]], align 8
9330 // CHECK18-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
9331 // CHECK18-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i64*
9332 // CHECK18-NEXT:    store i64 [[TMP71]], i64* [[TMP78]], align 8
9333 // CHECK18-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 0
9334 // CHECK18-NEXT:    store i8* null, i8** [[TMP79]], align 8
9335 // CHECK18-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 1
9336 // CHECK18-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
9337 // CHECK18-NEXT:    store i64 [[TMP73]], i64* [[TMP81]], align 8
9338 // CHECK18-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 1
9339 // CHECK18-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
9340 // CHECK18-NEXT:    store i64 [[TMP73]], i64* [[TMP83]], align 8
9341 // CHECK18-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS18]], i64 0, i64 1
9342 // CHECK18-NEXT:    store i8* null, i8** [[TMP84]], align 8
9343 // CHECK18-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS16]], i32 0, i32 0
9344 // CHECK18-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS17]], i32 0, i32 0
9345 // CHECK18-NEXT:    [[TMP87:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
9346 // CHECK18-NEXT:    [[TMP88:%.*]] = icmp ne i32 [[TMP87]], 0
9347 // CHECK18-NEXT:    br i1 [[TMP88]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
9348 // CHECK18:       omp_offload.failed19:
9349 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
9350 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT20]]
9351 // CHECK18:       omp_offload.cont20:
9352 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
9353 // CHECK18:       omp_if.else:
9354 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i64 [[TMP71]], i64 [[TMP73]]) #[[ATTR3]]
9355 // CHECK18-NEXT:    br label [[OMP_IF_END]]
9356 // CHECK18:       omp_if.end:
9357 // CHECK18-NEXT:    [[TMP89:%.*]] = load i32, i32* [[A]], align 4
9358 // CHECK18-NEXT:    [[CONV22:%.*]] = bitcast i64* [[A_CASTED21]] to i32*
9359 // CHECK18-NEXT:    store i32 [[TMP89]], i32* [[CONV22]], align 4
9360 // CHECK18-NEXT:    [[TMP90:%.*]] = load i64, i64* [[A_CASTED21]], align 8
9361 // CHECK18-NEXT:    [[TMP91:%.*]] = load i32, i32* [[N_ADDR]], align 4
9362 // CHECK18-NEXT:    [[CMP23:%.*]] = icmp sgt i32 [[TMP91]], 20
9363 // CHECK18-NEXT:    br i1 [[CMP23]], label [[OMP_IF_THEN24:%.*]], label [[OMP_IF_ELSE30:%.*]]
9364 // CHECK18:       omp_if.then24:
9365 // CHECK18-NEXT:    [[TMP92:%.*]] = mul nuw i64 [[TMP2]], 4
9366 // CHECK18-NEXT:    [[TMP93:%.*]] = mul nuw i64 5, [[TMP5]]
9367 // CHECK18-NEXT:    [[TMP94:%.*]] = mul nuw i64 [[TMP93]], 8
9368 // CHECK18-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
9369 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP95]], i8* align 8 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i64 72, i1 false)
9370 // CHECK18-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
9371 // CHECK18-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i64*
9372 // CHECK18-NEXT:    store i64 [[TMP90]], i64* [[TMP97]], align 8
9373 // CHECK18-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
9374 // CHECK18-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i64*
9375 // CHECK18-NEXT:    store i64 [[TMP90]], i64* [[TMP99]], align 8
9376 // CHECK18-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 0
9377 // CHECK18-NEXT:    store i8* null, i8** [[TMP100]], align 8
9378 // CHECK18-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 1
9379 // CHECK18-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
9380 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 8
9381 // CHECK18-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 1
9382 // CHECK18-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
9383 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 8
9384 // CHECK18-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 1
9385 // CHECK18-NEXT:    store i8* null, i8** [[TMP105]], align 8
9386 // CHECK18-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 2
9387 // CHECK18-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
9388 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP107]], align 8
9389 // CHECK18-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 2
9390 // CHECK18-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i64*
9391 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP109]], align 8
9392 // CHECK18-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 2
9393 // CHECK18-NEXT:    store i8* null, i8** [[TMP110]], align 8
9394 // CHECK18-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 3
9395 // CHECK18-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
9396 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP112]], align 8
9397 // CHECK18-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 3
9398 // CHECK18-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
9399 // CHECK18-NEXT:    store float* [[VLA]], float** [[TMP114]], align 8
9400 // CHECK18-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
9401 // CHECK18-NEXT:    store i64 [[TMP92]], i64* [[TMP115]], align 8
9402 // CHECK18-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 3
9403 // CHECK18-NEXT:    store i8* null, i8** [[TMP116]], align 8
9404 // CHECK18-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 4
9405 // CHECK18-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
9406 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 8
9407 // CHECK18-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 4
9408 // CHECK18-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
9409 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 8
9410 // CHECK18-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 4
9411 // CHECK18-NEXT:    store i8* null, i8** [[TMP121]], align 8
9412 // CHECK18-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 5
9413 // CHECK18-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i64*
9414 // CHECK18-NEXT:    store i64 5, i64* [[TMP123]], align 8
9415 // CHECK18-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 5
9416 // CHECK18-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i64*
9417 // CHECK18-NEXT:    store i64 5, i64* [[TMP125]], align 8
9418 // CHECK18-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 5
9419 // CHECK18-NEXT:    store i8* null, i8** [[TMP126]], align 8
9420 // CHECK18-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 6
9421 // CHECK18-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i64*
9422 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP128]], align 8
9423 // CHECK18-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 6
9424 // CHECK18-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i64*
9425 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP130]], align 8
9426 // CHECK18-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 6
9427 // CHECK18-NEXT:    store i8* null, i8** [[TMP131]], align 8
9428 // CHECK18-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 7
9429 // CHECK18-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
9430 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 8
9431 // CHECK18-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 7
9432 // CHECK18-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
9433 // CHECK18-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 8
9434 // CHECK18-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
9435 // CHECK18-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 8
9436 // CHECK18-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 7
9437 // CHECK18-NEXT:    store i8* null, i8** [[TMP137]], align 8
9438 // CHECK18-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 8
9439 // CHECK18-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
9440 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 8
9441 // CHECK18-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 8
9442 // CHECK18-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
9443 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 8
9444 // CHECK18-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS27]], i64 0, i64 8
9445 // CHECK18-NEXT:    store i8* null, i8** [[TMP142]], align 8
9446 // CHECK18-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS25]], i32 0, i32 0
9447 // CHECK18-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS26]], i32 0, i32 0
9448 // CHECK18-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9449 // CHECK18-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9450 // CHECK18-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
9451 // CHECK18-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
9452 // CHECK18:       omp_offload.failed28:
9453 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
9454 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT29]]
9455 // CHECK18:       omp_offload.cont29:
9456 // CHECK18-NEXT:    br label [[OMP_IF_END31:%.*]]
9457 // CHECK18:       omp_if.else30:
9458 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i64 [[TMP90]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
9459 // CHECK18-NEXT:    br label [[OMP_IF_END31]]
9460 // CHECK18:       omp_if.end31:
9461 // CHECK18-NEXT:    store i32 0, i32* [[NN]], align 4
9462 // CHECK18-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
9463 // CHECK18-NEXT:    [[CONV32:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
9464 // CHECK18-NEXT:    store i32 [[TMP148]], i32* [[CONV32]], align 4
9465 // CHECK18-NEXT:    [[TMP149:%.*]] = load i64, i64* [[NN_CASTED]], align 8
9466 // CHECK18-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
9467 // CHECK18-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
9468 // CHECK18-NEXT:    store i64 [[TMP149]], i64* [[TMP151]], align 8
9469 // CHECK18-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
9470 // CHECK18-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i64*
9471 // CHECK18-NEXT:    store i64 [[TMP149]], i64* [[TMP153]], align 8
9472 // CHECK18-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i64 0, i64 0
9473 // CHECK18-NEXT:    store i8* null, i8** [[TMP154]], align 8
9474 // CHECK18-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
9475 // CHECK18-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
9476 // CHECK18-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9477 // CHECK18-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
9478 // CHECK18-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
9479 // CHECK18:       omp_offload.failed36:
9480 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i64 [[TMP149]]) #[[ATTR3]]
9481 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
9482 // CHECK18:       omp_offload.cont37:
9483 // CHECK18-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
9484 // CHECK18-NEXT:    [[CONV39:%.*]] = bitcast i64* [[NN_CASTED38]] to i32*
9485 // CHECK18-NEXT:    store i32 [[TMP159]], i32* [[CONV39]], align 4
9486 // CHECK18-NEXT:    [[TMP160:%.*]] = load i64, i64* [[NN_CASTED38]], align 8
9487 // CHECK18-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
9488 // CHECK18-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i64*
9489 // CHECK18-NEXT:    store i64 [[TMP160]], i64* [[TMP162]], align 8
9490 // CHECK18-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
9491 // CHECK18-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i64*
9492 // CHECK18-NEXT:    store i64 [[TMP160]], i64* [[TMP164]], align 8
9493 // CHECK18-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS42]], i64 0, i64 0
9494 // CHECK18-NEXT:    store i8* null, i8** [[TMP165]], align 8
9495 // CHECK18-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS40]], i32 0, i32 0
9496 // CHECK18-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS41]], i32 0, i32 0
9497 // CHECK18-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9498 // CHECK18-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
9499 // CHECK18-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
9500 // CHECK18:       omp_offload.failed43:
9501 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i64 [[TMP160]]) #[[ATTR3]]
9502 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT44]]
9503 // CHECK18:       omp_offload.cont44:
9504 // CHECK18-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
9505 // CHECK18-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
9506 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
9507 // CHECK18-NEXT:    ret i32 [[TMP170]]
9508 //
9509 //
9510 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
9511 // CHECK18-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
9512 // CHECK18-NEXT:  entry:
9513 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9514 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9515 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
9516 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9517 // CHECK18-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
9518 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9519 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9520 // CHECK18-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
9521 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9522 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9523 // CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
9524 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
9525 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
9526 // CHECK18-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
9527 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
9528 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9529 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
9530 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9531 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
9532 // CHECK18-NEXT:    ret void
9533 //
9534 //
9535 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
9536 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
9537 // CHECK18-NEXT:  entry:
9538 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9539 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9540 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9541 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9542 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9543 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9544 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9545 // CHECK18-NEXT:    ret void
9546 //
9547 //
9548 // CHECK18-LABEL: define {{[^@]+}}@.omp_task_privates_map.
9549 // 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]+]] {
9550 // CHECK18-NEXT:  entry:
9551 // CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
9552 // CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 8
9553 // CHECK18-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 8
9554 // CHECK18-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 8
9555 // CHECK18-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 8
9556 // CHECK18-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
9557 // CHECK18-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 8
9558 // CHECK18-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 8
9559 // CHECK18-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 8
9560 // CHECK18-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 8
9561 // CHECK18-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
9562 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
9563 // CHECK18-NEXT:    [[TMP7:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 8
9564 // CHECK18-NEXT:    store [3 x i8*]* [[TMP6]], [3 x i8*]** [[TMP7]], align 8
9565 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
9566 // CHECK18-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 8
9567 // CHECK18-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 8
9568 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
9569 // CHECK18-NEXT:    [[TMP11:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 8
9570 // CHECK18-NEXT:    store [3 x i64]* [[TMP10]], [3 x i64]** [[TMP11]], align 8
9571 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
9572 // CHECK18-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 8
9573 // CHECK18-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 8
9574 // CHECK18-NEXT:    ret void
9575 //
9576 //
9577 // CHECK18-LABEL: define {{[^@]+}}@.omp_task_entry.
9578 // CHECK18-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
9579 // CHECK18-NEXT:  entry:
9580 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
9581 // CHECK18-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
9582 // CHECK18-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
9583 // CHECK18-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
9584 // CHECK18-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
9585 // CHECK18-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
9586 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 8
9587 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 8
9588 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 8
9589 // CHECK18-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 8
9590 // CHECK18-NEXT:    [[AA_CASTED_I:%.*]] = alloca i64, align 8
9591 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i64, align 8
9592 // CHECK18-NEXT:    [[DOTCAPTURE_EXPR__CASTED5_I:%.*]] = alloca i64, align 8
9593 // CHECK18-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
9594 // CHECK18-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
9595 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
9596 // CHECK18-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
9597 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
9598 // CHECK18-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
9599 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
9600 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
9601 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
9602 // CHECK18-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
9603 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
9604 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
9605 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
9606 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
9607 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META15:![0-9]+]])
9608 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META18:![0-9]+]])
9609 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
9610 // CHECK18-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
9611 // CHECK18-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !24
9612 // CHECK18-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !24
9613 // CHECK18-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
9614 // 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 !24
9615 // CHECK18-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !24
9616 // CHECK18-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
9617 // CHECK18-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !24
9618 // CHECK18-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !24
9619 // CHECK18-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !24
9620 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
9621 // 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]]
9622 // CHECK18-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !24
9623 // CHECK18-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 8, !noalias !24
9624 // CHECK18-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 8, !noalias !24
9625 // CHECK18-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 8, !noalias !24
9626 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i64 0, i64 0
9627 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i64 0, i64 0
9628 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i64 0, i64 0
9629 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
9630 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
9631 // CHECK18-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
9632 // CHECK18-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
9633 // CHECK18-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
9634 // CHECK18-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
9635 // CHECK18-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
9636 // CHECK18:       omp_offload.failed.i:
9637 // CHECK18-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
9638 // CHECK18-NEXT:    [[CONV_I:%.*]] = bitcast i64* [[AA_CASTED_I]] to i16*
9639 // CHECK18-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !24
9640 // CHECK18-NEXT:    [[TMP30:%.*]] = load i64, i64* [[AA_CASTED_I]], align 8, !noalias !24
9641 // CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
9642 // CHECK18-NEXT:    [[CONV4_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED_I]] to i32*
9643 // CHECK18-NEXT:    store i32 [[TMP31]], i32* [[CONV4_I]], align 4, !noalias !24
9644 // CHECK18-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED_I]], align 8, !noalias !24
9645 // CHECK18-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
9646 // CHECK18-NEXT:    [[CONV6_I:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED5_I]] to i32*
9647 // CHECK18-NEXT:    store i32 [[TMP33]], i32* [[CONV6_I]], align 4, !noalias !24
9648 // CHECK18-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED5_I]], align 8, !noalias !24
9649 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP30]], i64 [[TMP32]], i64 [[TMP34]]) #[[ATTR3]]
9650 // CHECK18-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
9651 // CHECK18:       .omp_outlined..1.exit:
9652 // CHECK18-NEXT:    ret i32 0
9653 //
9654 //
9655 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
9656 // CHECK18-SAME: (i64 noundef [[A:%.*]]) #[[ATTR2]] {
9657 // CHECK18-NEXT:  entry:
9658 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9659 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9660 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9661 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9662 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9663 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9664 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
9665 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
9666 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]])
9667 // CHECK18-NEXT:    ret void
9668 //
9669 //
9670 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
9671 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
9672 // CHECK18-NEXT:  entry:
9673 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9674 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9675 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9676 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9677 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9678 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9679 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9680 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9681 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
9682 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
9683 // CHECK18-NEXT:    ret void
9684 //
9685 //
9686 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
9687 // CHECK18-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR2]] {
9688 // CHECK18-NEXT:  entry:
9689 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9690 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9691 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9692 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9693 // CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
9694 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9695 // CHECK18-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
9696 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9697 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]])
9698 // CHECK18-NEXT:    ret void
9699 //
9700 //
9701 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
9702 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
9703 // CHECK18-NEXT:  entry:
9704 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9705 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9706 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9707 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9708 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9709 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9710 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9711 // CHECK18-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
9712 // CHECK18-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
9713 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
9714 // CHECK18-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
9715 // CHECK18-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
9716 // CHECK18-NEXT:    ret void
9717 //
9718 //
9719 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
9720 // CHECK18-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
9721 // CHECK18-NEXT:  entry:
9722 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9723 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9724 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9725 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9726 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9727 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9728 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9729 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9730 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9731 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9732 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
9733 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
9734 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
9735 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9736 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
9737 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9738 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
9739 // CHECK18-NEXT:    ret void
9740 //
9741 //
9742 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..6
9743 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
9744 // CHECK18-NEXT:  entry:
9745 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9746 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9747 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9748 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9749 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9750 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9751 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9752 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9753 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9754 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9755 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9756 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
9757 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
9758 // CHECK18-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
9759 // CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
9760 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
9761 // CHECK18-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
9762 // CHECK18-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
9763 // CHECK18-NEXT:    ret void
9764 //
9765 //
9766 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
9767 // 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:%.*]]) #[[ATTR2]] {
9768 // CHECK18-NEXT:  entry:
9769 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9770 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
9771 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9772 // CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
9773 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
9774 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9775 // CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
9776 // CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
9777 // CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
9778 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9779 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9780 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
9781 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9782 // CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
9783 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
9784 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9785 // CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
9786 // CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
9787 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
9788 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9789 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
9790 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9791 // CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
9792 // CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
9793 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9794 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
9795 // CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
9796 // CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
9797 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
9798 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9799 // CHECK18-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
9800 // CHECK18-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
9801 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
9802 // CHECK18-NEXT:    ret void
9803 //
9804 //
9805 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
9806 // 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:%.*]]) #[[ATTR2]] {
9807 // CHECK18-NEXT:  entry:
9808 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9809 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9810 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9811 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
9812 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9813 // CHECK18-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
9814 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
9815 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9816 // CHECK18-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
9817 // CHECK18-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
9818 // CHECK18-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
9819 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9820 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9821 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9822 // CHECK18-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
9823 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9824 // CHECK18-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
9825 // CHECK18-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
9826 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9827 // CHECK18-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
9828 // CHECK18-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
9829 // CHECK18-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
9830 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9831 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
9832 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9833 // CHECK18-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
9834 // CHECK18-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
9835 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9836 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
9837 // CHECK18-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
9838 // CHECK18-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
9839 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
9840 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
9841 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
9842 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
9843 // CHECK18-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
9844 // CHECK18-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
9845 // CHECK18-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
9846 // CHECK18-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
9847 // CHECK18-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
9848 // CHECK18-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
9849 // CHECK18-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
9850 // CHECK18-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
9851 // CHECK18-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
9852 // CHECK18-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
9853 // CHECK18-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
9854 // CHECK18-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
9855 // CHECK18-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
9856 // CHECK18-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
9857 // CHECK18-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
9858 // CHECK18-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
9859 // CHECK18-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
9860 // CHECK18-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
9861 // CHECK18-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
9862 // CHECK18-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
9863 // CHECK18-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
9864 // CHECK18-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
9865 // CHECK18-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
9866 // CHECK18-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
9867 // CHECK18-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
9868 // CHECK18-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
9869 // CHECK18-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
9870 // CHECK18-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
9871 // CHECK18-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
9872 // CHECK18-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
9873 // CHECK18-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
9874 // CHECK18-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
9875 // CHECK18-NEXT:    ret void
9876 //
9877 //
9878 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
9879 // CHECK18-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
9880 // CHECK18-NEXT:  entry:
9881 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
9882 // CHECK18-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
9883 // CHECK18-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
9884 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
9885 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9886 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
9887 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
9888 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
9889 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP1]])
9890 // CHECK18-NEXT:    ret void
9891 //
9892 //
9893 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..12
9894 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
9895 // CHECK18-NEXT:  entry:
9896 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9897 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9898 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
9899 // CHECK18-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
9900 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9901 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9902 // CHECK18-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
9903 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
9904 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9905 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
9906 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
9907 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
9908 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP1]])
9909 // CHECK18-NEXT:    ret void
9910 //
9911 //
9912 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..13
9913 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
9914 // CHECK18-NEXT:  entry:
9915 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9916 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9917 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
9918 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9919 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9920 // CHECK18-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
9921 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
9922 // CHECK18-NEXT:    ret void
9923 //
9924 //
9925 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
9926 // CHECK18-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR2]] {
9927 // CHECK18-NEXT:  entry:
9928 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
9929 // CHECK18-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
9930 // CHECK18-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
9931 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
9932 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
9933 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
9934 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
9935 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
9936 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP1]])
9937 // CHECK18-NEXT:    ret void
9938 //
9939 //
9940 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..16
9941 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR2]] {
9942 // CHECK18-NEXT:  entry:
9943 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9944 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9945 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
9946 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9947 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9948 // CHECK18-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
9949 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
9950 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[CONV]])
9951 // CHECK18-NEXT:    ret void
9952 //
9953 //
9954 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..17
9955 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
9956 // CHECK18-NEXT:  entry:
9957 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9958 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9959 // CHECK18-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
9960 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9961 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9962 // CHECK18-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
9963 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
9964 // CHECK18-NEXT:    ret void
9965 //
9966 //
9967 // CHECK18-LABEL: define {{[^@]+}}@_Z6bazzzziPi
9968 // CHECK18-SAME: (i32 noundef signext [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
9969 // CHECK18-NEXT:  entry:
9970 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
9971 // CHECK18-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 8
9972 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
9973 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
9974 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
9975 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
9976 // CHECK18-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 8
9977 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
9978 // CHECK18-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
9979 // CHECK18-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9980 // CHECK18-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
9981 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
9982 // CHECK18-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9983 // CHECK18-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
9984 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
9985 // CHECK18-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9986 // CHECK18-NEXT:    store i8* null, i8** [[TMP6]], align 8
9987 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9988 // CHECK18-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9989 // CHECK18-NEXT:    [[TMP9:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP7]], i8** [[TMP8]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9990 // CHECK18-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
9991 // CHECK18-NEXT:    br i1 [[TMP10]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9992 // CHECK18:       omp_offload.failed:
9993 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i64 [[TMP1]]) #[[ATTR3]]
9994 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9995 // CHECK18:       omp_offload.cont:
9996 // CHECK18-NEXT:    ret void
9997 //
9998 //
9999 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
10000 // CHECK18-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
10001 // CHECK18-NEXT:  entry:
10002 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10003 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10004 // CHECK18-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10005 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i64 [[TMP0]])
10006 // CHECK18-NEXT:    ret void
10007 //
10008 //
10009 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..20
10010 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR2]] {
10011 // CHECK18-NEXT:  entry:
10012 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10013 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10014 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10015 // CHECK18-NEXT:    [[F:%.*]] = alloca i32*, align 8
10016 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10017 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10018 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10019 // CHECK18-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10020 // CHECK18-NEXT:    ret void
10021 //
10022 //
10023 // CHECK18-LABEL: define {{[^@]+}}@_Z3bari
10024 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
10025 // CHECK18-NEXT:  entry:
10026 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10027 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
10028 // CHECK18-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
10029 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10030 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
10031 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
10032 // CHECK18-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
10033 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
10034 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
10035 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
10036 // CHECK18-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
10037 // CHECK18-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
10038 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
10039 // CHECK18-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
10040 // CHECK18-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
10041 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
10042 // CHECK18-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
10043 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
10044 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
10045 // CHECK18-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
10046 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
10047 // CHECK18-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
10048 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
10049 // CHECK18-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
10050 // CHECK18-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
10051 // CHECK18-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
10052 // CHECK18-NEXT:    ret i32 [[TMP8]]
10053 //
10054 //
10055 // CHECK18-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
10056 // CHECK18-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
10057 // CHECK18-NEXT:  entry:
10058 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10059 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10060 // CHECK18-NEXT:    [[B:%.*]] = alloca i32, align 4
10061 // CHECK18-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
10062 // CHECK18-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
10063 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
10064 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
10065 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
10066 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
10067 // CHECK18-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
10068 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10069 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10070 // CHECK18-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10071 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
10072 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
10073 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
10074 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
10075 // CHECK18-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
10076 // CHECK18-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
10077 // CHECK18-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
10078 // CHECK18-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
10079 // CHECK18-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
10080 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
10081 // CHECK18-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
10082 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
10083 // CHECK18-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
10084 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
10085 // CHECK18-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
10086 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
10087 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10088 // CHECK18:       omp_if.then:
10089 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
10090 // CHECK18-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
10091 // CHECK18-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
10092 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
10093 // CHECK18-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i64 40, i1 false)
10094 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10095 // CHECK18-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
10096 // CHECK18-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
10097 // CHECK18-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10098 // CHECK18-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
10099 // CHECK18-NEXT:    store double* [[A]], double** [[TMP14]], align 8
10100 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10101 // CHECK18-NEXT:    store i8* null, i8** [[TMP15]], align 8
10102 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10103 // CHECK18-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
10104 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
10105 // CHECK18-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10106 // CHECK18-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
10107 // CHECK18-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
10108 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
10109 // CHECK18-NEXT:    store i8* null, i8** [[TMP20]], align 8
10110 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10111 // CHECK18-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
10112 // CHECK18-NEXT:    store i64 2, i64* [[TMP22]], align 8
10113 // CHECK18-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10114 // CHECK18-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
10115 // CHECK18-NEXT:    store i64 2, i64* [[TMP24]], align 8
10116 // CHECK18-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
10117 // CHECK18-NEXT:    store i8* null, i8** [[TMP25]], align 8
10118 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
10119 // CHECK18-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
10120 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
10121 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
10122 // CHECK18-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
10123 // CHECK18-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
10124 // CHECK18-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
10125 // CHECK18-NEXT:    store i8* null, i8** [[TMP30]], align 8
10126 // CHECK18-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
10127 // CHECK18-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
10128 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
10129 // CHECK18-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
10130 // CHECK18-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
10131 // CHECK18-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
10132 // CHECK18-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
10133 // CHECK18-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
10134 // CHECK18-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
10135 // CHECK18-NEXT:    store i8* null, i8** [[TMP36]], align 8
10136 // CHECK18-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10137 // CHECK18-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10138 // CHECK18-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
10139 // CHECK18-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10140 // CHECK18-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
10141 // CHECK18-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10142 // CHECK18:       omp_offload.failed:
10143 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
10144 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10145 // CHECK18:       omp_offload.cont:
10146 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
10147 // CHECK18:       omp_if.else:
10148 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR3]]
10149 // CHECK18-NEXT:    br label [[OMP_IF_END]]
10150 // CHECK18:       omp_if.end:
10151 // CHECK18-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
10152 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
10153 // CHECK18-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
10154 // CHECK18-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
10155 // CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
10156 // CHECK18-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
10157 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
10158 // CHECK18-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
10159 // CHECK18-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
10160 // CHECK18-NEXT:    ret i32 [[ADD4]]
10161 //
10162 //
10163 // CHECK18-LABEL: define {{[^@]+}}@_ZL7fstatici
10164 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
10165 // CHECK18-NEXT:  entry:
10166 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10167 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
10168 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
10169 // CHECK18-NEXT:    [[AAA:%.*]] = alloca i8, align 1
10170 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
10171 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10172 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10173 // CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
10174 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
10175 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
10176 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
10177 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10178 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
10179 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
10180 // CHECK18-NEXT:    store i8 0, i8* [[AAA]], align 1
10181 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
10182 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10183 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
10184 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
10185 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
10186 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10187 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
10188 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10189 // CHECK18-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
10190 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
10191 // CHECK18-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
10192 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
10193 // CHECK18-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
10194 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
10195 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10196 // CHECK18:       omp_if.then:
10197 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10198 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
10199 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
10200 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10201 // CHECK18-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
10202 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
10203 // CHECK18-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10204 // CHECK18-NEXT:    store i8* null, i8** [[TMP11]], align 8
10205 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10206 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
10207 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
10208 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10209 // CHECK18-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
10210 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
10211 // CHECK18-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
10212 // CHECK18-NEXT:    store i8* null, i8** [[TMP16]], align 8
10213 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10214 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
10215 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
10216 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10217 // CHECK18-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
10218 // CHECK18-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
10219 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
10220 // CHECK18-NEXT:    store i8* null, i8** [[TMP21]], align 8
10221 // CHECK18-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
10222 // CHECK18-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
10223 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
10224 // CHECK18-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
10225 // CHECK18-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
10226 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
10227 // CHECK18-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
10228 // CHECK18-NEXT:    store i8* null, i8** [[TMP26]], align 8
10229 // CHECK18-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10230 // CHECK18-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10231 // CHECK18-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10232 // CHECK18-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
10233 // CHECK18-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10234 // CHECK18:       omp_offload.failed:
10235 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
10236 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10237 // CHECK18:       omp_offload.cont:
10238 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
10239 // CHECK18:       omp_if.else:
10240 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
10241 // CHECK18-NEXT:    br label [[OMP_IF_END]]
10242 // CHECK18:       omp_if.end:
10243 // CHECK18-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
10244 // CHECK18-NEXT:    ret i32 [[TMP31]]
10245 //
10246 //
10247 // CHECK18-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
10248 // CHECK18-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
10249 // CHECK18-NEXT:  entry:
10250 // CHECK18-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10251 // CHECK18-NEXT:    [[A:%.*]] = alloca i32, align 4
10252 // CHECK18-NEXT:    [[AA:%.*]] = alloca i16, align 2
10253 // CHECK18-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
10254 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10255 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10256 // CHECK18-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
10257 // CHECK18-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
10258 // CHECK18-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
10259 // CHECK18-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10260 // CHECK18-NEXT:    store i32 0, i32* [[A]], align 4
10261 // CHECK18-NEXT:    store i16 0, i16* [[AA]], align 2
10262 // CHECK18-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
10263 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10264 // CHECK18-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
10265 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
10266 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
10267 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10268 // CHECK18-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
10269 // CHECK18-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10270 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
10271 // CHECK18-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
10272 // CHECK18-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10273 // CHECK18:       omp_if.then:
10274 // CHECK18-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10275 // CHECK18-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
10276 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
10277 // CHECK18-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10278 // CHECK18-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
10279 // CHECK18-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
10280 // CHECK18-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10281 // CHECK18-NEXT:    store i8* null, i8** [[TMP9]], align 8
10282 // CHECK18-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10283 // CHECK18-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
10284 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
10285 // CHECK18-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10286 // CHECK18-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
10287 // CHECK18-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
10288 // CHECK18-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
10289 // CHECK18-NEXT:    store i8* null, i8** [[TMP14]], align 8
10290 // CHECK18-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10291 // CHECK18-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
10292 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
10293 // CHECK18-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10294 // CHECK18-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
10295 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
10296 // CHECK18-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
10297 // CHECK18-NEXT:    store i8* null, i8** [[TMP19]], align 8
10298 // CHECK18-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10299 // CHECK18-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10300 // CHECK18-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10301 // CHECK18-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
10302 // CHECK18-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10303 // CHECK18:       omp_offload.failed:
10304 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
10305 // CHECK18-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10306 // CHECK18:       omp_offload.cont:
10307 // CHECK18-NEXT:    br label [[OMP_IF_END:%.*]]
10308 // CHECK18:       omp_if.else:
10309 // CHECK18-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
10310 // CHECK18-NEXT:    br label [[OMP_IF_END]]
10311 // CHECK18:       omp_if.end:
10312 // CHECK18-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
10313 // CHECK18-NEXT:    ret i32 [[TMP24]]
10314 //
10315 //
10316 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
10317 // CHECK18-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
10318 // CHECK18-NEXT:  entry:
10319 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10320 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
10321 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10322 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10323 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
10324 // CHECK18-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
10325 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10326 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
10327 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10328 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10329 // CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
10330 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10331 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
10332 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10333 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10334 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
10335 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
10336 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
10337 // CHECK18-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
10338 // CHECK18-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
10339 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
10340 // CHECK18-NEXT:    ret void
10341 //
10342 //
10343 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..23
10344 // 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]] {
10345 // CHECK18-NEXT:  entry:
10346 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10347 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10348 // CHECK18-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10349 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
10350 // CHECK18-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10351 // CHECK18-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10352 // CHECK18-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
10353 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10354 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10355 // CHECK18-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10356 // CHECK18-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
10357 // CHECK18-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10358 // CHECK18-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10359 // CHECK18-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
10360 // CHECK18-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10361 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
10362 // CHECK18-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10363 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10364 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
10365 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
10366 // CHECK18-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
10367 // CHECK18-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
10368 // CHECK18-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
10369 // CHECK18-NEXT:    store double [[ADD]], double* [[A]], align 8
10370 // CHECK18-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
10371 // CHECK18-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
10372 // CHECK18-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
10373 // CHECK18-NEXT:    store double [[INC]], double* [[A4]], align 8
10374 // CHECK18-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
10375 // CHECK18-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
10376 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
10377 // CHECK18-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
10378 // CHECK18-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
10379 // CHECK18-NEXT:    ret void
10380 //
10381 //
10382 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
10383 // CHECK18-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
10384 // CHECK18-NEXT:  entry:
10385 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10386 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10387 // CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
10388 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10389 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10390 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10391 // CHECK18-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
10392 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10393 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10394 // CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
10395 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10396 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10397 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10398 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
10399 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10400 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
10401 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10402 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
10403 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
10404 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
10405 // CHECK18-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10406 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
10407 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10408 // CHECK18-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
10409 // CHECK18-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
10410 // CHECK18-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
10411 // CHECK18-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
10412 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
10413 // CHECK18-NEXT:    ret void
10414 //
10415 //
10416 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..26
10417 // CHECK18-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
10418 // CHECK18-NEXT:  entry:
10419 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10420 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10421 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10422 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10423 // CHECK18-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
10424 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10425 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10426 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10427 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10428 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10429 // CHECK18-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
10430 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10431 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10432 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10433 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
10434 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10435 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
10436 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
10437 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
10438 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
10439 // CHECK18-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
10440 // CHECK18-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
10441 // CHECK18-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
10442 // CHECK18-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
10443 // CHECK18-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
10444 // CHECK18-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
10445 // CHECK18-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
10446 // CHECK18-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
10447 // CHECK18-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
10448 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
10449 // CHECK18-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10450 // CHECK18-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
10451 // CHECK18-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
10452 // CHECK18-NEXT:    ret void
10453 //
10454 //
10455 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
10456 // CHECK18-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
10457 // CHECK18-NEXT:  entry:
10458 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10459 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10460 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10461 // CHECK18-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10462 // CHECK18-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10463 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10464 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10465 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10466 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10467 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10468 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10469 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
10470 // CHECK18-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10471 // CHECK18-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
10472 // CHECK18-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
10473 // CHECK18-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
10474 // CHECK18-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10475 // CHECK18-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
10476 // CHECK18-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10477 // CHECK18-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
10478 // CHECK18-NEXT:    ret void
10479 //
10480 //
10481 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..29
10482 // 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]] {
10483 // CHECK18-NEXT:  entry:
10484 // CHECK18-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10485 // CHECK18-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10486 // CHECK18-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10487 // CHECK18-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10488 // CHECK18-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10489 // CHECK18-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10490 // CHECK18-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10491 // CHECK18-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10492 // CHECK18-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10493 // CHECK18-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10494 // CHECK18-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10495 // CHECK18-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10496 // CHECK18-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10497 // CHECK18-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
10498 // CHECK18-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
10499 // CHECK18-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
10500 // CHECK18-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
10501 // CHECK18-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
10502 // CHECK18-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
10503 // CHECK18-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
10504 // CHECK18-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
10505 // CHECK18-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
10506 // CHECK18-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
10507 // CHECK18-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
10508 // CHECK18-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
10509 // CHECK18-NEXT:    ret void
10510 //
10511 //
10512 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
10513 // CHECK18-SAME: () #[[ATTR4]] {
10514 // CHECK18-NEXT:  entry:
10515 // CHECK18-NEXT:    call void @__tgt_register_requires(i64 1)
10516 // CHECK18-NEXT:    ret void
10517 //
10518 //
10519 // CHECK19-LABEL: define {{[^@]+}}@_Z3fooi
10520 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
10521 // CHECK19-NEXT:  entry:
10522 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
10523 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
10524 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
10525 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
10526 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
10527 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
10528 // CHECK19-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
10529 // CHECK19-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
10530 // CHECK19-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
10531 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10532 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
10533 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10534 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
10535 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
10536 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
10537 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
10538 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
10539 // CHECK19-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
10540 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10541 // CHECK19-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
10542 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
10543 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
10544 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
10545 // CHECK19-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
10546 // CHECK19-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
10547 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
10548 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
10549 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
10550 // CHECK19-NEXT:    [[A_CASTED17:%.*]] = alloca i32, align 4
10551 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [9 x i8*], align 4
10552 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [9 x i8*], align 4
10553 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [9 x i8*], align 4
10554 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
10555 // CHECK19-NEXT:    [[NN:%.*]] = alloca i32, align 4
10556 // CHECK19-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
10557 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
10558 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
10559 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
10560 // CHECK19-NEXT:    [[NN_CASTED32:%.*]] = alloca i32, align 4
10561 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 4
10562 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 4
10563 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 4
10564 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
10565 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
10566 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
10567 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
10568 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
10569 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
10570 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
10571 // CHECK19-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
10572 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
10573 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
10574 // CHECK19-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
10575 // CHECK19-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
10576 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
10577 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
10578 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
10579 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
10580 // CHECK19-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
10581 // CHECK19-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
10582 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10583 // CHECK19-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
10584 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10585 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10586 // CHECK19-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
10587 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
10588 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10589 // CHECK19-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
10590 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
10591 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10592 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
10593 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
10594 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10595 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
10596 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
10597 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
10598 // CHECK19-NEXT:    store i8* null, i8** [[TMP17]], align 4
10599 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
10600 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
10601 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
10602 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
10603 // CHECK19-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
10604 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
10605 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
10606 // CHECK19-NEXT:    store i8* null, i8** [[TMP22]], align 4
10607 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
10608 // CHECK19-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
10609 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
10610 // CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
10611 // CHECK19-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
10612 // CHECK19-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
10613 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
10614 // CHECK19-NEXT:    store i8* null, i8** [[TMP27]], align 4
10615 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10616 // CHECK19-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10617 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
10618 // CHECK19-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
10619 // CHECK19-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
10620 // CHECK19-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
10621 // CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10622 // CHECK19-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
10623 // CHECK19-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
10624 // CHECK19-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10625 // CHECK19-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
10626 // CHECK19-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
10627 // CHECK19-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
10628 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
10629 // CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
10630 // CHECK19-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
10631 // CHECK19-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
10632 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
10633 // CHECK19-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
10634 // CHECK19-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
10635 // CHECK19-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
10636 // CHECK19-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
10637 // 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)
10638 // CHECK19-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
10639 // CHECK19-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
10640 // CHECK19-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
10641 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
10642 // CHECK19-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
10643 // CHECK19-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
10644 // CHECK19-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
10645 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
10646 // CHECK19-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
10647 // CHECK19-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
10648 // CHECK19-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
10649 // CHECK19-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP36]])
10650 // CHECK19-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
10651 // CHECK19-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
10652 // CHECK19-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
10653 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
10654 // CHECK19-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
10655 // CHECK19-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
10656 // CHECK19-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
10657 // CHECK19-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
10658 // CHECK19-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
10659 // CHECK19-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
10660 // CHECK19-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
10661 // CHECK19-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
10662 // CHECK19-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
10663 // CHECK19-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
10664 // CHECK19-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
10665 // CHECK19-NEXT:    store i8* null, i8** [[TMP63]], align 4
10666 // CHECK19-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
10667 // CHECK19-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
10668 // CHECK19-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
10669 // CHECK19-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
10670 // CHECK19-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10671 // CHECK19:       omp_offload.failed:
10672 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i32 [[TMP58]]) #[[ATTR3]]
10673 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10674 // CHECK19:       omp_offload.cont:
10675 // CHECK19-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
10676 // CHECK19-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
10677 // CHECK19-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
10678 // CHECK19-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
10679 // CHECK19-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
10680 // CHECK19-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
10681 // CHECK19-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
10682 // CHECK19-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
10683 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
10684 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10685 // CHECK19:       omp_if.then:
10686 // CHECK19-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
10687 // CHECK19-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
10688 // CHECK19-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
10689 // CHECK19-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
10690 // CHECK19-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
10691 // CHECK19-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
10692 // CHECK19-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
10693 // CHECK19-NEXT:    store i8* null, i8** [[TMP77]], align 4
10694 // CHECK19-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
10695 // CHECK19-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
10696 // CHECK19-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
10697 // CHECK19-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
10698 // CHECK19-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
10699 // CHECK19-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
10700 // CHECK19-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
10701 // CHECK19-NEXT:    store i8* null, i8** [[TMP82]], align 4
10702 // CHECK19-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
10703 // CHECK19-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
10704 // CHECK19-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
10705 // CHECK19-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
10706 // CHECK19-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
10707 // CHECK19:       omp_offload.failed15:
10708 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
10709 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
10710 // CHECK19:       omp_offload.cont16:
10711 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
10712 // CHECK19:       omp_if.else:
10713 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
10714 // CHECK19-NEXT:    br label [[OMP_IF_END]]
10715 // CHECK19:       omp_if.end:
10716 // CHECK19-NEXT:    [[TMP87:%.*]] = load i32, i32* [[A]], align 4
10717 // CHECK19-NEXT:    store i32 [[TMP87]], i32* [[A_CASTED17]], align 4
10718 // CHECK19-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A_CASTED17]], align 4
10719 // CHECK19-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
10720 // CHECK19-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP89]], 20
10721 // CHECK19-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
10722 // CHECK19:       omp_if.then19:
10723 // CHECK19-NEXT:    [[TMP90:%.*]] = mul nuw i32 [[TMP1]], 4
10724 // CHECK19-NEXT:    [[TMP91:%.*]] = sext i32 [[TMP90]] to i64
10725 // CHECK19-NEXT:    [[TMP92:%.*]] = mul nuw i32 5, [[TMP3]]
10726 // CHECK19-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP92]], 8
10727 // CHECK19-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
10728 // CHECK19-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
10729 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP95]], i8* align 4 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i32 72, i1 false)
10730 // CHECK19-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
10731 // CHECK19-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i32*
10732 // CHECK19-NEXT:    store i32 [[TMP88]], i32* [[TMP97]], align 4
10733 // CHECK19-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
10734 // CHECK19-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
10735 // CHECK19-NEXT:    store i32 [[TMP88]], i32* [[TMP99]], align 4
10736 // CHECK19-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
10737 // CHECK19-NEXT:    store i8* null, i8** [[TMP100]], align 4
10738 // CHECK19-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
10739 // CHECK19-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
10740 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 4
10741 // CHECK19-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
10742 // CHECK19-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
10743 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 4
10744 // CHECK19-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
10745 // CHECK19-NEXT:    store i8* null, i8** [[TMP105]], align 4
10746 // CHECK19-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
10747 // CHECK19-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i32*
10748 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP107]], align 4
10749 // CHECK19-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
10750 // CHECK19-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i32*
10751 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP109]], align 4
10752 // CHECK19-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 2
10753 // CHECK19-NEXT:    store i8* null, i8** [[TMP110]], align 4
10754 // CHECK19-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
10755 // CHECK19-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
10756 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP112]], align 4
10757 // CHECK19-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
10758 // CHECK19-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
10759 // CHECK19-NEXT:    store float* [[VLA]], float** [[TMP114]], align 4
10760 // CHECK19-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
10761 // CHECK19-NEXT:    store i64 [[TMP91]], i64* [[TMP115]], align 4
10762 // CHECK19-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 3
10763 // CHECK19-NEXT:    store i8* null, i8** [[TMP116]], align 4
10764 // CHECK19-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
10765 // CHECK19-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
10766 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 4
10767 // CHECK19-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
10768 // CHECK19-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
10769 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 4
10770 // CHECK19-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 4
10771 // CHECK19-NEXT:    store i8* null, i8** [[TMP121]], align 4
10772 // CHECK19-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
10773 // CHECK19-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i32*
10774 // CHECK19-NEXT:    store i32 5, i32* [[TMP123]], align 4
10775 // CHECK19-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
10776 // CHECK19-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i32*
10777 // CHECK19-NEXT:    store i32 5, i32* [[TMP125]], align 4
10778 // CHECK19-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 5
10779 // CHECK19-NEXT:    store i8* null, i8** [[TMP126]], align 4
10780 // CHECK19-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
10781 // CHECK19-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
10782 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP128]], align 4
10783 // CHECK19-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
10784 // CHECK19-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i32*
10785 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP130]], align 4
10786 // CHECK19-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 6
10787 // CHECK19-NEXT:    store i8* null, i8** [[TMP131]], align 4
10788 // CHECK19-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
10789 // CHECK19-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
10790 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 4
10791 // CHECK19-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
10792 // CHECK19-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
10793 // CHECK19-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 4
10794 // CHECK19-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
10795 // CHECK19-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 4
10796 // CHECK19-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 7
10797 // CHECK19-NEXT:    store i8* null, i8** [[TMP137]], align 4
10798 // CHECK19-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
10799 // CHECK19-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
10800 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 4
10801 // CHECK19-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
10802 // CHECK19-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
10803 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 4
10804 // CHECK19-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 8
10805 // CHECK19-NEXT:    store i8* null, i8** [[TMP142]], align 4
10806 // CHECK19-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
10807 // CHECK19-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
10808 // CHECK19-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
10809 // CHECK19-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10810 // CHECK19-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
10811 // CHECK19-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
10812 // CHECK19:       omp_offload.failed23:
10813 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
10814 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
10815 // CHECK19:       omp_offload.cont24:
10816 // CHECK19-NEXT:    br label [[OMP_IF_END26:%.*]]
10817 // CHECK19:       omp_if.else25:
10818 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
10819 // CHECK19-NEXT:    br label [[OMP_IF_END26]]
10820 // CHECK19:       omp_if.end26:
10821 // CHECK19-NEXT:    store i32 0, i32* [[NN]], align 4
10822 // CHECK19-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
10823 // CHECK19-NEXT:    store i32 [[TMP148]], i32* [[NN_CASTED]], align 4
10824 // CHECK19-NEXT:    [[TMP149:%.*]] = load i32, i32* [[NN_CASTED]], align 4
10825 // CHECK19-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
10826 // CHECK19-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i32*
10827 // CHECK19-NEXT:    store i32 [[TMP149]], i32* [[TMP151]], align 4
10828 // CHECK19-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
10829 // CHECK19-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
10830 // CHECK19-NEXT:    store i32 [[TMP149]], i32* [[TMP153]], align 4
10831 // CHECK19-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
10832 // CHECK19-NEXT:    store i8* null, i8** [[TMP154]], align 4
10833 // CHECK19-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
10834 // CHECK19-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
10835 // CHECK19-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10836 // CHECK19-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
10837 // CHECK19-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED30:%.*]], label [[OMP_OFFLOAD_CONT31:%.*]]
10838 // CHECK19:       omp_offload.failed30:
10839 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i32 [[TMP149]]) #[[ATTR3]]
10840 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT31]]
10841 // CHECK19:       omp_offload.cont31:
10842 // CHECK19-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
10843 // CHECK19-NEXT:    store i32 [[TMP159]], i32* [[NN_CASTED32]], align 4
10844 // CHECK19-NEXT:    [[TMP160:%.*]] = load i32, i32* [[NN_CASTED32]], align 4
10845 // CHECK19-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
10846 // CHECK19-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i32*
10847 // CHECK19-NEXT:    store i32 [[TMP160]], i32* [[TMP162]], align 4
10848 // CHECK19-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
10849 // CHECK19-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i32*
10850 // CHECK19-NEXT:    store i32 [[TMP160]], i32* [[TMP164]], align 4
10851 // CHECK19-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i32 0, i32 0
10852 // CHECK19-NEXT:    store i8* null, i8** [[TMP165]], align 4
10853 // CHECK19-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
10854 // CHECK19-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
10855 // CHECK19-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10856 // CHECK19-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
10857 // CHECK19-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
10858 // CHECK19:       omp_offload.failed36:
10859 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i32 [[TMP160]]) #[[ATTR3]]
10860 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
10861 // CHECK19:       omp_offload.cont37:
10862 // CHECK19-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
10863 // CHECK19-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
10864 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
10865 // CHECK19-NEXT:    ret i32 [[TMP170]]
10866 //
10867 //
10868 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
10869 // CHECK19-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
10870 // CHECK19-NEXT:  entry:
10871 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10872 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
10873 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
10874 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10875 // CHECK19-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
10876 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10877 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
10878 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
10879 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10880 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
10881 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
10882 // CHECK19-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
10883 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
10884 // CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10885 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
10886 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10887 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
10888 // CHECK19-NEXT:    ret void
10889 //
10890 //
10891 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
10892 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
10893 // CHECK19-NEXT:  entry:
10894 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10895 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10896 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10897 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10898 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10899 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10900 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10901 // CHECK19-NEXT:    ret void
10902 //
10903 //
10904 // CHECK19-LABEL: define {{[^@]+}}@.omp_task_privates_map.
10905 // 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]+]] {
10906 // CHECK19-NEXT:  entry:
10907 // CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
10908 // CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
10909 // CHECK19-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
10910 // CHECK19-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
10911 // CHECK19-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
10912 // CHECK19-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
10913 // CHECK19-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
10914 // CHECK19-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
10915 // CHECK19-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
10916 // CHECK19-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
10917 // CHECK19-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
10918 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
10919 // CHECK19-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
10920 // CHECK19-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
10921 // CHECK19-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
10922 // CHECK19-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
10923 // CHECK19-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
10924 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
10925 // CHECK19-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
10926 // CHECK19-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
10927 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
10928 // CHECK19-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
10929 // CHECK19-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
10930 // CHECK19-NEXT:    ret void
10931 //
10932 //
10933 // CHECK19-LABEL: define {{[^@]+}}@.omp_task_entry.
10934 // CHECK19-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
10935 // CHECK19-NEXT:  entry:
10936 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
10937 // CHECK19-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
10938 // CHECK19-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
10939 // CHECK19-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
10940 // CHECK19-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
10941 // CHECK19-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
10942 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
10943 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
10944 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
10945 // CHECK19-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
10946 // CHECK19-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
10947 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
10948 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
10949 // CHECK19-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
10950 // CHECK19-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
10951 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
10952 // CHECK19-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
10953 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
10954 // CHECK19-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
10955 // CHECK19-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
10956 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
10957 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
10958 // CHECK19-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
10959 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
10960 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
10961 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
10962 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
10963 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
10964 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
10965 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
10966 // CHECK19-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
10967 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
10968 // CHECK19-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !25
10969 // CHECK19-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
10970 // 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 !25
10971 // CHECK19-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !25
10972 // CHECK19-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
10973 // CHECK19-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
10974 // CHECK19-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !25
10975 // CHECK19-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
10976 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
10977 // 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]]
10978 // CHECK19-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !25
10979 // CHECK19-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !25
10980 // CHECK19-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !25
10981 // CHECK19-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !25
10982 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
10983 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
10984 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
10985 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
10986 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
10987 // CHECK19-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
10988 // CHECK19-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
10989 // CHECK19-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
10990 // CHECK19-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
10991 // CHECK19-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
10992 // CHECK19:       omp_offload.failed.i:
10993 // CHECK19-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
10994 // CHECK19-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
10995 // CHECK19-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !25
10996 // CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !25
10997 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
10998 // CHECK19-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
10999 // CHECK19-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
11000 // CHECK19-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
11001 // CHECK19-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
11002 // CHECK19-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
11003 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
11004 // CHECK19-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
11005 // CHECK19:       .omp_outlined..1.exit:
11006 // CHECK19-NEXT:    ret i32 0
11007 //
11008 //
11009 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
11010 // CHECK19-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
11011 // CHECK19-NEXT:  entry:
11012 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11013 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11014 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11015 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
11016 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
11017 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
11018 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
11019 // CHECK19-NEXT:    ret void
11020 //
11021 //
11022 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
11023 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
11024 // CHECK19-NEXT:  entry:
11025 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11026 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11027 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11028 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11029 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11030 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11031 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
11032 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
11033 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
11034 // CHECK19-NEXT:    ret void
11035 //
11036 //
11037 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
11038 // CHECK19-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
11039 // CHECK19-NEXT:  entry:
11040 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11041 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11042 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11043 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11044 // CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
11045 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11046 // CHECK19-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
11047 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11048 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
11049 // CHECK19-NEXT:    ret void
11050 //
11051 //
11052 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
11053 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
11054 // CHECK19-NEXT:  entry:
11055 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11056 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11057 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11058 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11059 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11060 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11061 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11062 // CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
11063 // CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
11064 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
11065 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
11066 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
11067 // CHECK19-NEXT:    ret void
11068 //
11069 //
11070 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
11071 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
11072 // CHECK19-NEXT:  entry:
11073 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11074 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11075 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11076 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11077 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11078 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11079 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11080 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
11081 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
11082 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
11083 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
11084 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11085 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
11086 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11087 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
11088 // CHECK19-NEXT:    ret void
11089 //
11090 //
11091 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
11092 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
11093 // CHECK19-NEXT:  entry:
11094 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11095 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11096 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11097 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11098 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11099 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11100 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11101 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11102 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11103 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
11104 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
11105 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
11106 // CHECK19-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
11107 // CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
11108 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
11109 // CHECK19-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
11110 // CHECK19-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
11111 // CHECK19-NEXT:    ret void
11112 //
11113 //
11114 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
11115 // 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:%.*]]) #[[ATTR2]] {
11116 // CHECK19-NEXT:  entry:
11117 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11118 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11119 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11120 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11121 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11122 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11123 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11124 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11125 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11126 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11127 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11128 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
11129 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11130 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
11131 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
11132 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11133 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
11134 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
11135 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
11136 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
11137 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11138 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
11139 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
11140 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11141 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
11142 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
11143 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
11144 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
11145 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
11146 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
11147 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
11148 // CHECK19-NEXT:    ret void
11149 //
11150 //
11151 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
11152 // 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:%.*]]) #[[ATTR2]] {
11153 // CHECK19-NEXT:  entry:
11154 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11155 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11156 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11157 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11158 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11159 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11160 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11161 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11162 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11163 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11164 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11165 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11166 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11167 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11168 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
11169 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11170 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
11171 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
11172 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11173 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
11174 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
11175 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
11176 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
11177 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11178 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
11179 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
11180 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11181 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
11182 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
11183 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
11184 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
11185 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
11186 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
11187 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
11188 // CHECK19-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
11189 // CHECK19-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
11190 // CHECK19-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
11191 // CHECK19-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
11192 // CHECK19-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
11193 // CHECK19-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
11194 // CHECK19-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
11195 // CHECK19-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
11196 // CHECK19-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
11197 // CHECK19-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
11198 // CHECK19-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
11199 // CHECK19-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
11200 // CHECK19-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
11201 // CHECK19-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
11202 // CHECK19-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
11203 // CHECK19-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
11204 // CHECK19-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
11205 // CHECK19-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
11206 // CHECK19-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
11207 // CHECK19-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
11208 // CHECK19-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
11209 // CHECK19-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
11210 // CHECK19-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
11211 // CHECK19-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
11212 // CHECK19-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
11213 // CHECK19-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
11214 // CHECK19-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
11215 // CHECK19-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
11216 // CHECK19-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
11217 // CHECK19-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
11218 // CHECK19-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
11219 // CHECK19-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
11220 // CHECK19-NEXT:    ret void
11221 //
11222 //
11223 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
11224 // CHECK19-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
11225 // CHECK19-NEXT:  entry:
11226 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
11227 // CHECK19-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
11228 // CHECK19-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
11229 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
11230 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
11231 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
11232 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP1]])
11233 // CHECK19-NEXT:    ret void
11234 //
11235 //
11236 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..12
11237 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
11238 // CHECK19-NEXT:  entry:
11239 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11240 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11241 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
11242 // CHECK19-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
11243 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11244 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11245 // CHECK19-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
11246 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
11247 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
11248 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
11249 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP1]])
11250 // CHECK19-NEXT:    ret void
11251 //
11252 //
11253 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..13
11254 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
11255 // CHECK19-NEXT:  entry:
11256 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11257 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11258 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
11259 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11260 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11261 // CHECK19-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
11262 // CHECK19-NEXT:    ret void
11263 //
11264 //
11265 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
11266 // CHECK19-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
11267 // CHECK19-NEXT:  entry:
11268 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
11269 // CHECK19-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
11270 // CHECK19-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
11271 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
11272 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
11273 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
11274 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP1]])
11275 // CHECK19-NEXT:    ret void
11276 //
11277 //
11278 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..16
11279 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
11280 // CHECK19-NEXT:  entry:
11281 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11282 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11283 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
11284 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11285 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11286 // CHECK19-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
11287 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
11288 // CHECK19-NEXT:    ret void
11289 //
11290 //
11291 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..17
11292 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
11293 // CHECK19-NEXT:  entry:
11294 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11295 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11296 // CHECK19-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
11297 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11298 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11299 // CHECK19-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
11300 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
11301 // CHECK19-NEXT:    ret void
11302 //
11303 //
11304 // CHECK19-LABEL: define {{[^@]+}}@_Z6bazzzziPi
11305 // CHECK19-SAME: (i32 noundef [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
11306 // CHECK19-NEXT:  entry:
11307 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11308 // CHECK19-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 4
11309 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
11310 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
11311 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
11312 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11313 // CHECK19-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 4
11314 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11315 // CHECK19-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11316 // CHECK19-NEXT:    [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32*
11317 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP2]], align 4
11318 // CHECK19-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11319 // CHECK19-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
11320 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[TMP4]], align 4
11321 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
11322 // CHECK19-NEXT:    store i8* null, i8** [[TMP5]], align 4
11323 // CHECK19-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11324 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11325 // CHECK19-NEXT:    [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11326 // CHECK19-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
11327 // CHECK19-NEXT:    br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11328 // CHECK19:       omp_offload.failed:
11329 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i32 [[TMP0]]) #[[ATTR3]]
11330 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11331 // CHECK19:       omp_offload.cont:
11332 // CHECK19-NEXT:    ret void
11333 //
11334 //
11335 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
11336 // CHECK19-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
11337 // CHECK19-NEXT:  entry:
11338 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11339 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11340 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11341 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i32 [[TMP0]])
11342 // CHECK19-NEXT:    ret void
11343 //
11344 //
11345 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..20
11346 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
11347 // CHECK19-NEXT:  entry:
11348 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11349 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11350 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11351 // CHECK19-NEXT:    [[F:%.*]] = alloca i32*, align 4
11352 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11353 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11354 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11355 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11356 // CHECK19-NEXT:    ret void
11357 //
11358 //
11359 // CHECK19-LABEL: define {{[^@]+}}@_Z3bari
11360 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
11361 // CHECK19-NEXT:  entry:
11362 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11363 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
11364 // CHECK19-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
11365 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11366 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
11367 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11368 // CHECK19-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
11369 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
11370 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
11371 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
11372 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
11373 // CHECK19-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
11374 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
11375 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
11376 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
11377 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
11378 // CHECK19-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
11379 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11380 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
11381 // CHECK19-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
11382 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
11383 // CHECK19-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
11384 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
11385 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
11386 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
11387 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
11388 // CHECK19-NEXT:    ret i32 [[TMP8]]
11389 //
11390 //
11391 // CHECK19-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
11392 // CHECK19-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
11393 // CHECK19-NEXT:  entry:
11394 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11395 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11396 // CHECK19-NEXT:    [[B:%.*]] = alloca i32, align 4
11397 // CHECK19-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
11398 // CHECK19-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
11399 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
11400 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
11401 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
11402 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
11403 // CHECK19-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
11404 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11405 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11406 // CHECK19-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11407 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
11408 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
11409 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
11410 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
11411 // CHECK19-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
11412 // CHECK19-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
11413 // CHECK19-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
11414 // CHECK19-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
11415 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
11416 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
11417 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
11418 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
11419 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
11420 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
11421 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11422 // CHECK19:       omp_if.then:
11423 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
11424 // CHECK19-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
11425 // CHECK19-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
11426 // CHECK19-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
11427 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
11428 // CHECK19-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i32 40, i1 false)
11429 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11430 // CHECK19-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
11431 // CHECK19-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
11432 // CHECK19-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11433 // CHECK19-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
11434 // CHECK19-NEXT:    store double* [[A]], double** [[TMP14]], align 4
11435 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
11436 // CHECK19-NEXT:    store i8* null, i8** [[TMP15]], align 4
11437 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11438 // CHECK19-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
11439 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
11440 // CHECK19-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11441 // CHECK19-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
11442 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
11443 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
11444 // CHECK19-NEXT:    store i8* null, i8** [[TMP20]], align 4
11445 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11446 // CHECK19-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
11447 // CHECK19-NEXT:    store i32 2, i32* [[TMP22]], align 4
11448 // CHECK19-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11449 // CHECK19-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
11450 // CHECK19-NEXT:    store i32 2, i32* [[TMP24]], align 4
11451 // CHECK19-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
11452 // CHECK19-NEXT:    store i8* null, i8** [[TMP25]], align 4
11453 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
11454 // CHECK19-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
11455 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
11456 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
11457 // CHECK19-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
11458 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
11459 // CHECK19-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
11460 // CHECK19-NEXT:    store i8* null, i8** [[TMP30]], align 4
11461 // CHECK19-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
11462 // CHECK19-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
11463 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
11464 // CHECK19-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
11465 // CHECK19-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
11466 // CHECK19-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
11467 // CHECK19-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
11468 // CHECK19-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
11469 // CHECK19-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
11470 // CHECK19-NEXT:    store i8* null, i8** [[TMP36]], align 4
11471 // CHECK19-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11472 // CHECK19-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11473 // CHECK19-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
11474 // CHECK19-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11475 // CHECK19-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
11476 // CHECK19-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11477 // CHECK19:       omp_offload.failed:
11478 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
11479 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11480 // CHECK19:       omp_offload.cont:
11481 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
11482 // CHECK19:       omp_if.else:
11483 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
11484 // CHECK19-NEXT:    br label [[OMP_IF_END]]
11485 // CHECK19:       omp_if.end:
11486 // CHECK19-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
11487 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
11488 // CHECK19-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
11489 // CHECK19-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
11490 // CHECK19-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
11491 // CHECK19-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
11492 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
11493 // CHECK19-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
11494 // CHECK19-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
11495 // CHECK19-NEXT:    ret i32 [[ADD3]]
11496 //
11497 //
11498 // CHECK19-LABEL: define {{[^@]+}}@_ZL7fstatici
11499 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
11500 // CHECK19-NEXT:  entry:
11501 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11502 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
11503 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
11504 // CHECK19-NEXT:    [[AAA:%.*]] = alloca i8, align 1
11505 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
11506 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11507 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11508 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
11509 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
11510 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
11511 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
11512 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11513 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
11514 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
11515 // CHECK19-NEXT:    store i8 0, i8* [[AAA]], align 1
11516 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
11517 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
11518 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
11519 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
11520 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11521 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
11522 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11523 // CHECK19-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
11524 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
11525 // CHECK19-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
11526 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
11527 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
11528 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
11529 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11530 // CHECK19:       omp_if.then:
11531 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11532 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
11533 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
11534 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11535 // CHECK19-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
11536 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
11537 // CHECK19-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
11538 // CHECK19-NEXT:    store i8* null, i8** [[TMP11]], align 4
11539 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11540 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
11541 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
11542 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11543 // CHECK19-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
11544 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
11545 // CHECK19-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
11546 // CHECK19-NEXT:    store i8* null, i8** [[TMP16]], align 4
11547 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11548 // CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
11549 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
11550 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11551 // CHECK19-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
11552 // CHECK19-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
11553 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
11554 // CHECK19-NEXT:    store i8* null, i8** [[TMP21]], align 4
11555 // CHECK19-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
11556 // CHECK19-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
11557 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
11558 // CHECK19-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
11559 // CHECK19-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
11560 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
11561 // CHECK19-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
11562 // CHECK19-NEXT:    store i8* null, i8** [[TMP26]], align 4
11563 // CHECK19-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11564 // CHECK19-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11565 // CHECK19-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11566 // CHECK19-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
11567 // CHECK19-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11568 // CHECK19:       omp_offload.failed:
11569 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
11570 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11571 // CHECK19:       omp_offload.cont:
11572 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
11573 // CHECK19:       omp_if.else:
11574 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
11575 // CHECK19-NEXT:    br label [[OMP_IF_END]]
11576 // CHECK19:       omp_if.end:
11577 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
11578 // CHECK19-NEXT:    ret i32 [[TMP31]]
11579 //
11580 //
11581 // CHECK19-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
11582 // CHECK19-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
11583 // CHECK19-NEXT:  entry:
11584 // CHECK19-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11585 // CHECK19-NEXT:    [[A:%.*]] = alloca i32, align 4
11586 // CHECK19-NEXT:    [[AA:%.*]] = alloca i16, align 2
11587 // CHECK19-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
11588 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11589 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11590 // CHECK19-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
11591 // CHECK19-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
11592 // CHECK19-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
11593 // CHECK19-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11594 // CHECK19-NEXT:    store i32 0, i32* [[A]], align 4
11595 // CHECK19-NEXT:    store i16 0, i16* [[AA]], align 2
11596 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
11597 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
11598 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
11599 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
11600 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11601 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
11602 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11603 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
11604 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
11605 // CHECK19-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11606 // CHECK19:       omp_if.then:
11607 // CHECK19-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11608 // CHECK19-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
11609 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
11610 // CHECK19-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11611 // CHECK19-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
11612 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
11613 // CHECK19-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
11614 // CHECK19-NEXT:    store i8* null, i8** [[TMP9]], align 4
11615 // CHECK19-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11616 // CHECK19-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
11617 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
11618 // CHECK19-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11619 // CHECK19-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
11620 // CHECK19-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
11621 // CHECK19-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
11622 // CHECK19-NEXT:    store i8* null, i8** [[TMP14]], align 4
11623 // CHECK19-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11624 // CHECK19-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
11625 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
11626 // CHECK19-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11627 // CHECK19-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
11628 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
11629 // CHECK19-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
11630 // CHECK19-NEXT:    store i8* null, i8** [[TMP19]], align 4
11631 // CHECK19-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11632 // CHECK19-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11633 // CHECK19-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11634 // CHECK19-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
11635 // CHECK19-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11636 // CHECK19:       omp_offload.failed:
11637 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
11638 // CHECK19-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11639 // CHECK19:       omp_offload.cont:
11640 // CHECK19-NEXT:    br label [[OMP_IF_END:%.*]]
11641 // CHECK19:       omp_if.else:
11642 // CHECK19-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
11643 // CHECK19-NEXT:    br label [[OMP_IF_END]]
11644 // CHECK19:       omp_if.end:
11645 // CHECK19-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
11646 // CHECK19-NEXT:    ret i32 [[TMP24]]
11647 //
11648 //
11649 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
11650 // CHECK19-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
11651 // CHECK19-NEXT:  entry:
11652 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11653 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
11654 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11655 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11656 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
11657 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
11658 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11659 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
11660 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11661 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11662 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
11663 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11664 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11665 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11666 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
11667 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
11668 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
11669 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
11670 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
11671 // CHECK19-NEXT:    ret void
11672 //
11673 //
11674 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..23
11675 // 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]] {
11676 // CHECK19-NEXT:  entry:
11677 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11678 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11679 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11680 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
11681 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11682 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11683 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
11684 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11685 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11686 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11687 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
11688 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11689 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11690 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
11691 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11692 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11693 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11694 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
11695 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
11696 // CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
11697 // CHECK19-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
11698 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
11699 // CHECK19-NEXT:    store double [[ADD]], double* [[A]], align 4
11700 // CHECK19-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11701 // CHECK19-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
11702 // CHECK19-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
11703 // CHECK19-NEXT:    store double [[INC]], double* [[A3]], align 4
11704 // CHECK19-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
11705 // CHECK19-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
11706 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
11707 // CHECK19-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
11708 // CHECK19-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
11709 // CHECK19-NEXT:    ret void
11710 //
11711 //
11712 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
11713 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11714 // CHECK19-NEXT:  entry:
11715 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11716 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11717 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
11718 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11719 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11720 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11721 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
11722 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11723 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11724 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
11725 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11726 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11727 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
11728 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11729 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11730 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
11731 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
11732 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
11733 // CHECK19-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11734 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
11735 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11736 // CHECK19-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
11737 // CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
11738 // CHECK19-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
11739 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
11740 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
11741 // CHECK19-NEXT:    ret void
11742 //
11743 //
11744 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..26
11745 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11746 // CHECK19-NEXT:  entry:
11747 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11748 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11749 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11750 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11751 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
11752 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11753 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11754 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11755 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11756 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11757 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
11758 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11759 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11760 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
11761 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11762 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11763 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
11764 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
11765 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
11766 // CHECK19-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
11767 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
11768 // CHECK19-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
11769 // CHECK19-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
11770 // CHECK19-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
11771 // CHECK19-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
11772 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
11773 // CHECK19-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
11774 // CHECK19-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
11775 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
11776 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11777 // CHECK19-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
11778 // CHECK19-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
11779 // CHECK19-NEXT:    ret void
11780 //
11781 //
11782 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
11783 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
11784 // CHECK19-NEXT:  entry:
11785 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11786 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11787 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11788 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11789 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11790 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11791 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11792 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11793 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11794 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11795 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11796 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
11797 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
11798 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
11799 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11800 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
11801 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11802 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
11803 // CHECK19-NEXT:    ret void
11804 //
11805 //
11806 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..29
11807 // 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]] {
11808 // CHECK19-NEXT:  entry:
11809 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11810 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11811 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11812 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11813 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11814 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11815 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11816 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11817 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11818 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11819 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11820 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11821 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11822 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
11823 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
11824 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
11825 // CHECK19-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
11826 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
11827 // CHECK19-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
11828 // CHECK19-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
11829 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
11830 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
11831 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
11832 // CHECK19-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
11833 // CHECK19-NEXT:    ret void
11834 //
11835 //
11836 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
11837 // CHECK19-SAME: () #[[ATTR4]] {
11838 // CHECK19-NEXT:  entry:
11839 // CHECK19-NEXT:    call void @__tgt_register_requires(i64 1)
11840 // CHECK19-NEXT:    ret void
11841 //
11842 //
11843 // CHECK20-LABEL: define {{[^@]+}}@_Z3fooi
11844 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
11845 // CHECK20-NEXT:  entry:
11846 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
11847 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
11848 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
11849 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
11850 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
11851 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
11852 // CHECK20-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
11853 // CHECK20-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
11854 // CHECK20-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
11855 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11856 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
11857 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11858 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
11859 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED3:%.*]] = alloca i32, align 4
11860 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
11861 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
11862 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
11863 // CHECK20-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 4
11864 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11865 // CHECK20-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
11866 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [1 x i8*], align 4
11867 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [1 x i8*], align 4
11868 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [1 x i8*], align 4
11869 // CHECK20-NEXT:    [[A_CASTED9:%.*]] = alloca i32, align 4
11870 // CHECK20-NEXT:    [[AA_CASTED10:%.*]] = alloca i32, align 4
11871 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS12:%.*]] = alloca [2 x i8*], align 4
11872 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS13:%.*]] = alloca [2 x i8*], align 4
11873 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS14:%.*]] = alloca [2 x i8*], align 4
11874 // CHECK20-NEXT:    [[A_CASTED17:%.*]] = alloca i32, align 4
11875 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [9 x i8*], align 4
11876 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [9 x i8*], align 4
11877 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [9 x i8*], align 4
11878 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [9 x i64], align 4
11879 // CHECK20-NEXT:    [[NN:%.*]] = alloca i32, align 4
11880 // CHECK20-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
11881 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS27:%.*]] = alloca [1 x i8*], align 4
11882 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS28:%.*]] = alloca [1 x i8*], align 4
11883 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS29:%.*]] = alloca [1 x i8*], align 4
11884 // CHECK20-NEXT:    [[NN_CASTED32:%.*]] = alloca i32, align 4
11885 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS33:%.*]] = alloca [1 x i8*], align 4
11886 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS34:%.*]] = alloca [1 x i8*], align 4
11887 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS35:%.*]] = alloca [1 x i8*], align 4
11888 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
11889 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
11890 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
11891 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
11892 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
11893 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
11894 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
11895 // CHECK20-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
11896 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
11897 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
11898 // CHECK20-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
11899 // CHECK20-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
11900 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
11901 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
11902 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_]], align 4
11903 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[A]], align 4
11904 // CHECK20-NEXT:    store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_2]], align 4
11905 // CHECK20-NEXT:    [[TMP7:%.*]] = load i16, i16* [[AA]], align 2
11906 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11907 // CHECK20-NEXT:    store i16 [[TMP7]], i16* [[CONV]], align 2
11908 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11909 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11910 // CHECK20-NEXT:    store i32 [[TMP9]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
11911 // CHECK20-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
11912 // CHECK20-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11913 // CHECK20-NEXT:    store i32 [[TMP11]], i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
11914 // CHECK20-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED3]], align 4
11915 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11916 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i32*
11917 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP14]], align 4
11918 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11919 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i32*
11920 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[TMP16]], align 4
11921 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
11922 // CHECK20-NEXT:    store i8* null, i8** [[TMP17]], align 4
11923 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
11924 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
11925 // CHECK20-NEXT:    store i32 [[TMP10]], i32* [[TMP19]], align 4
11926 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
11927 // CHECK20-NEXT:    [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
11928 // CHECK20-NEXT:    store i32 [[TMP10]], i32* [[TMP21]], align 4
11929 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
11930 // CHECK20-NEXT:    store i8* null, i8** [[TMP22]], align 4
11931 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
11932 // CHECK20-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
11933 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP24]], align 4
11934 // CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
11935 // CHECK20-NEXT:    [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i32*
11936 // CHECK20-NEXT:    store i32 [[TMP12]], i32* [[TMP26]], align 4
11937 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
11938 // CHECK20-NEXT:    store i8* null, i8** [[TMP27]], align 4
11939 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11940 // CHECK20-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11941 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
11942 // CHECK20-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2
11943 // CHECK20-NEXT:    store i16 [[TMP31]], i16* [[TMP30]], align 4
11944 // CHECK20-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
11945 // CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11946 // CHECK20-NEXT:    store i32 [[TMP33]], i32* [[TMP32]], align 4
11947 // CHECK20-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
11948 // CHECK20-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11949 // CHECK20-NEXT:    store i32 [[TMP35]], i32* [[TMP34]], align 4
11950 // CHECK20-NEXT:    [[TMP36:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB1]], 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)
11951 // CHECK20-NEXT:    [[TMP37:%.*]] = bitcast i8* [[TMP36]] to %struct.kmp_task_t_with_privates*
11952 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 0
11953 // CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP38]], i32 0, i32 0
11954 // CHECK20-NEXT:    [[TMP40:%.*]] = load i8*, i8** [[TMP39]], align 4
11955 // CHECK20-NEXT:    [[TMP41:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
11956 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP40]], i8* align 4 [[TMP41]], i32 12, i1 false)
11957 // CHECK20-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP37]], i32 0, i32 1
11958 // CHECK20-NEXT:    [[TMP43:%.*]] = bitcast i8* [[TMP40]] to %struct.anon*
11959 // CHECK20-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 0
11960 // CHECK20-NEXT:    [[TMP45:%.*]] = bitcast [3 x i64]* [[TMP44]] to i8*
11961 // 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)
11962 // CHECK20-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 1
11963 // CHECK20-NEXT:    [[TMP47:%.*]] = bitcast [3 x i8*]* [[TMP46]] to i8*
11964 // CHECK20-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP28]] to i8*
11965 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP47]], i8* align 4 [[TMP48]], i32 12, i1 false)
11966 // CHECK20-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 2
11967 // CHECK20-NEXT:    [[TMP50:%.*]] = bitcast [3 x i8*]* [[TMP49]] to i8*
11968 // CHECK20-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP29]] to i8*
11969 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP50]], i8* align 4 [[TMP51]], i32 12, i1 false)
11970 // CHECK20-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP42]], i32 0, i32 3
11971 // CHECK20-NEXT:    [[TMP53:%.*]] = load i16, i16* [[AA]], align 2
11972 // CHECK20-NEXT:    store i16 [[TMP53]], i16* [[TMP52]], align 4
11973 // CHECK20-NEXT:    [[TMP54:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i8* [[TMP36]])
11974 // CHECK20-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
11975 // CHECK20-NEXT:    store i32 [[TMP55]], i32* [[A_CASTED]], align 4
11976 // CHECK20-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A_CASTED]], align 4
11977 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105(i32 [[TMP56]]) #[[ATTR3:[0-9]+]]
11978 // CHECK20-NEXT:    [[TMP57:%.*]] = load i16, i16* [[AA]], align 2
11979 // CHECK20-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
11980 // CHECK20-NEXT:    store i16 [[TMP57]], i16* [[CONV5]], align 2
11981 // CHECK20-NEXT:    [[TMP58:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
11982 // CHECK20-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
11983 // CHECK20-NEXT:    [[TMP60:%.*]] = bitcast i8** [[TMP59]] to i32*
11984 // CHECK20-NEXT:    store i32 [[TMP58]], i32* [[TMP60]], align 4
11985 // CHECK20-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
11986 // CHECK20-NEXT:    [[TMP62:%.*]] = bitcast i8** [[TMP61]] to i32*
11987 // CHECK20-NEXT:    store i32 [[TMP58]], i32* [[TMP62]], align 4
11988 // CHECK20-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
11989 // CHECK20-NEXT:    store i8* null, i8** [[TMP63]], align 4
11990 // CHECK20-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
11991 // CHECK20-NEXT:    [[TMP65:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
11992 // CHECK20-NEXT:    [[TMP66:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111.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)
11993 // CHECK20-NEXT:    [[TMP67:%.*]] = icmp ne i32 [[TMP66]], 0
11994 // CHECK20-NEXT:    br i1 [[TMP67]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11995 // CHECK20:       omp_offload.failed:
11996 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111(i32 [[TMP58]]) #[[ATTR3]]
11997 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11998 // CHECK20:       omp_offload.cont:
11999 // CHECK20-NEXT:    [[TMP68:%.*]] = load i32, i32* [[A]], align 4
12000 // CHECK20-NEXT:    store i32 [[TMP68]], i32* [[A_CASTED9]], align 4
12001 // CHECK20-NEXT:    [[TMP69:%.*]] = load i32, i32* [[A_CASTED9]], align 4
12002 // CHECK20-NEXT:    [[TMP70:%.*]] = load i16, i16* [[AA]], align 2
12003 // CHECK20-NEXT:    [[CONV11:%.*]] = bitcast i32* [[AA_CASTED10]] to i16*
12004 // CHECK20-NEXT:    store i16 [[TMP70]], i16* [[CONV11]], align 2
12005 // CHECK20-NEXT:    [[TMP71:%.*]] = load i32, i32* [[AA_CASTED10]], align 4
12006 // CHECK20-NEXT:    [[TMP72:%.*]] = load i32, i32* [[N_ADDR]], align 4
12007 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP72]], 10
12008 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12009 // CHECK20:       omp_if.then:
12010 // CHECK20-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
12011 // CHECK20-NEXT:    [[TMP74:%.*]] = bitcast i8** [[TMP73]] to i32*
12012 // CHECK20-NEXT:    store i32 [[TMP69]], i32* [[TMP74]], align 4
12013 // CHECK20-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
12014 // CHECK20-NEXT:    [[TMP76:%.*]] = bitcast i8** [[TMP75]] to i32*
12015 // CHECK20-NEXT:    store i32 [[TMP69]], i32* [[TMP76]], align 4
12016 // CHECK20-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 0
12017 // CHECK20-NEXT:    store i8* null, i8** [[TMP77]], align 4
12018 // CHECK20-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 1
12019 // CHECK20-NEXT:    [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
12020 // CHECK20-NEXT:    store i32 [[TMP71]], i32* [[TMP79]], align 4
12021 // CHECK20-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 1
12022 // CHECK20-NEXT:    [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i32*
12023 // CHECK20-NEXT:    store i32 [[TMP71]], i32* [[TMP81]], align 4
12024 // CHECK20-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS14]], i32 0, i32 1
12025 // CHECK20-NEXT:    store i8* null, i8** [[TMP82]], align 4
12026 // CHECK20-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS12]], i32 0, i32 0
12027 // CHECK20-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS13]], i32 0, i32 0
12028 // CHECK20-NEXT:    [[TMP85:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118.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)
12029 // CHECK20-NEXT:    [[TMP86:%.*]] = icmp ne i32 [[TMP85]], 0
12030 // CHECK20-NEXT:    br i1 [[TMP86]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
12031 // CHECK20:       omp_offload.failed15:
12032 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
12033 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT16]]
12034 // CHECK20:       omp_offload.cont16:
12035 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
12036 // CHECK20:       omp_if.else:
12037 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118(i32 [[TMP69]], i32 [[TMP71]]) #[[ATTR3]]
12038 // CHECK20-NEXT:    br label [[OMP_IF_END]]
12039 // CHECK20:       omp_if.end:
12040 // CHECK20-NEXT:    [[TMP87:%.*]] = load i32, i32* [[A]], align 4
12041 // CHECK20-NEXT:    store i32 [[TMP87]], i32* [[A_CASTED17]], align 4
12042 // CHECK20-NEXT:    [[TMP88:%.*]] = load i32, i32* [[A_CASTED17]], align 4
12043 // CHECK20-NEXT:    [[TMP89:%.*]] = load i32, i32* [[N_ADDR]], align 4
12044 // CHECK20-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP89]], 20
12045 // CHECK20-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
12046 // CHECK20:       omp_if.then19:
12047 // CHECK20-NEXT:    [[TMP90:%.*]] = mul nuw i32 [[TMP1]], 4
12048 // CHECK20-NEXT:    [[TMP91:%.*]] = sext i32 [[TMP90]] to i64
12049 // CHECK20-NEXT:    [[TMP92:%.*]] = mul nuw i32 5, [[TMP3]]
12050 // CHECK20-NEXT:    [[TMP93:%.*]] = mul nuw i32 [[TMP92]], 8
12051 // CHECK20-NEXT:    [[TMP94:%.*]] = sext i32 [[TMP93]] to i64
12052 // CHECK20-NEXT:    [[TMP95:%.*]] = bitcast [9 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
12053 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP95]], i8* align 4 bitcast ([9 x i64]* @.offload_sizes.10 to i8*), i32 72, i1 false)
12054 // CHECK20-NEXT:    [[TMP96:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
12055 // CHECK20-NEXT:    [[TMP97:%.*]] = bitcast i8** [[TMP96]] to i32*
12056 // CHECK20-NEXT:    store i32 [[TMP88]], i32* [[TMP97]], align 4
12057 // CHECK20-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
12058 // CHECK20-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
12059 // CHECK20-NEXT:    store i32 [[TMP88]], i32* [[TMP99]], align 4
12060 // CHECK20-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
12061 // CHECK20-NEXT:    store i8* null, i8** [[TMP100]], align 4
12062 // CHECK20-NEXT:    [[TMP101:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
12063 // CHECK20-NEXT:    [[TMP102:%.*]] = bitcast i8** [[TMP101]] to [10 x float]**
12064 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP102]], align 4
12065 // CHECK20-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
12066 // CHECK20-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to [10 x float]**
12067 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP104]], align 4
12068 // CHECK20-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
12069 // CHECK20-NEXT:    store i8* null, i8** [[TMP105]], align 4
12070 // CHECK20-NEXT:    [[TMP106:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
12071 // CHECK20-NEXT:    [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i32*
12072 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP107]], align 4
12073 // CHECK20-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
12074 // CHECK20-NEXT:    [[TMP109:%.*]] = bitcast i8** [[TMP108]] to i32*
12075 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP109]], align 4
12076 // CHECK20-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 2
12077 // CHECK20-NEXT:    store i8* null, i8** [[TMP110]], align 4
12078 // CHECK20-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
12079 // CHECK20-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to float**
12080 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP112]], align 4
12081 // CHECK20-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
12082 // CHECK20-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to float**
12083 // CHECK20-NEXT:    store float* [[VLA]], float** [[TMP114]], align 4
12084 // CHECK20-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
12085 // CHECK20-NEXT:    store i64 [[TMP91]], i64* [[TMP115]], align 4
12086 // CHECK20-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 3
12087 // CHECK20-NEXT:    store i8* null, i8** [[TMP116]], align 4
12088 // CHECK20-NEXT:    [[TMP117:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
12089 // CHECK20-NEXT:    [[TMP118:%.*]] = bitcast i8** [[TMP117]] to [5 x [10 x double]]**
12090 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP118]], align 4
12091 // CHECK20-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
12092 // CHECK20-NEXT:    [[TMP120:%.*]] = bitcast i8** [[TMP119]] to [5 x [10 x double]]**
12093 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP120]], align 4
12094 // CHECK20-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 4
12095 // CHECK20-NEXT:    store i8* null, i8** [[TMP121]], align 4
12096 // CHECK20-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
12097 // CHECK20-NEXT:    [[TMP123:%.*]] = bitcast i8** [[TMP122]] to i32*
12098 // CHECK20-NEXT:    store i32 5, i32* [[TMP123]], align 4
12099 // CHECK20-NEXT:    [[TMP124:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
12100 // CHECK20-NEXT:    [[TMP125:%.*]] = bitcast i8** [[TMP124]] to i32*
12101 // CHECK20-NEXT:    store i32 5, i32* [[TMP125]], align 4
12102 // CHECK20-NEXT:    [[TMP126:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 5
12103 // CHECK20-NEXT:    store i8* null, i8** [[TMP126]], align 4
12104 // CHECK20-NEXT:    [[TMP127:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
12105 // CHECK20-NEXT:    [[TMP128:%.*]] = bitcast i8** [[TMP127]] to i32*
12106 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP128]], align 4
12107 // CHECK20-NEXT:    [[TMP129:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
12108 // CHECK20-NEXT:    [[TMP130:%.*]] = bitcast i8** [[TMP129]] to i32*
12109 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP130]], align 4
12110 // CHECK20-NEXT:    [[TMP131:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 6
12111 // CHECK20-NEXT:    store i8* null, i8** [[TMP131]], align 4
12112 // CHECK20-NEXT:    [[TMP132:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
12113 // CHECK20-NEXT:    [[TMP133:%.*]] = bitcast i8** [[TMP132]] to double**
12114 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP133]], align 4
12115 // CHECK20-NEXT:    [[TMP134:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
12116 // CHECK20-NEXT:    [[TMP135:%.*]] = bitcast i8** [[TMP134]] to double**
12117 // CHECK20-NEXT:    store double* [[VLA1]], double** [[TMP135]], align 4
12118 // CHECK20-NEXT:    [[TMP136:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
12119 // CHECK20-NEXT:    store i64 [[TMP94]], i64* [[TMP136]], align 4
12120 // CHECK20-NEXT:    [[TMP137:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 7
12121 // CHECK20-NEXT:    store i8* null, i8** [[TMP137]], align 4
12122 // CHECK20-NEXT:    [[TMP138:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
12123 // CHECK20-NEXT:    [[TMP139:%.*]] = bitcast i8** [[TMP138]] to %struct.TT**
12124 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP139]], align 4
12125 // CHECK20-NEXT:    [[TMP140:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
12126 // CHECK20-NEXT:    [[TMP141:%.*]] = bitcast i8** [[TMP140]] to %struct.TT**
12127 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP141]], align 4
12128 // CHECK20-NEXT:    [[TMP142:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 8
12129 // CHECK20-NEXT:    store i8* null, i8** [[TMP142]], align 4
12130 // CHECK20-NEXT:    [[TMP143:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
12131 // CHECK20-NEXT:    [[TMP144:%.*]] = getelementptr inbounds [9 x i8*], [9 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
12132 // CHECK20-NEXT:    [[TMP145:%.*]] = getelementptr inbounds [9 x i64], [9 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
12133 // CHECK20-NEXT:    [[TMP146:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142.region_id, i32 9, i8** [[TMP143]], i8** [[TMP144]], i64* [[TMP145]], i64* getelementptr inbounds ([9 x i64], [9 x i64]* @.offload_maptypes.11, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12134 // CHECK20-NEXT:    [[TMP147:%.*]] = icmp ne i32 [[TMP146]], 0
12135 // CHECK20-NEXT:    br i1 [[TMP147]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
12136 // CHECK20:       omp_offload.failed23:
12137 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
12138 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
12139 // CHECK20:       omp_offload.cont24:
12140 // CHECK20-NEXT:    br label [[OMP_IF_END26:%.*]]
12141 // CHECK20:       omp_if.else25:
12142 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142(i32 [[TMP88]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]]) #[[ATTR3]]
12143 // CHECK20-NEXT:    br label [[OMP_IF_END26]]
12144 // CHECK20:       omp_if.end26:
12145 // CHECK20-NEXT:    store i32 0, i32* [[NN]], align 4
12146 // CHECK20-NEXT:    [[TMP148:%.*]] = load i32, i32* [[NN]], align 4
12147 // CHECK20-NEXT:    store i32 [[TMP148]], i32* [[NN_CASTED]], align 4
12148 // CHECK20-NEXT:    [[TMP149:%.*]] = load i32, i32* [[NN_CASTED]], align 4
12149 // CHECK20-NEXT:    [[TMP150:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
12150 // CHECK20-NEXT:    [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i32*
12151 // CHECK20-NEXT:    store i32 [[TMP149]], i32* [[TMP151]], align 4
12152 // CHECK20-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
12153 // CHECK20-NEXT:    [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
12154 // CHECK20-NEXT:    store i32 [[TMP149]], i32* [[TMP153]], align 4
12155 // CHECK20-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS29]], i32 0, i32 0
12156 // CHECK20-NEXT:    store i8* null, i8** [[TMP154]], align 4
12157 // CHECK20-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS27]], i32 0, i32 0
12158 // CHECK20-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS28]], i32 0, i32 0
12159 // CHECK20-NEXT:    [[TMP157:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154.region_id, i32 1, i8** [[TMP155]], i8** [[TMP156]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12160 // CHECK20-NEXT:    [[TMP158:%.*]] = icmp ne i32 [[TMP157]], 0
12161 // CHECK20-NEXT:    br i1 [[TMP158]], label [[OMP_OFFLOAD_FAILED30:%.*]], label [[OMP_OFFLOAD_CONT31:%.*]]
12162 // CHECK20:       omp_offload.failed30:
12163 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154(i32 [[TMP149]]) #[[ATTR3]]
12164 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT31]]
12165 // CHECK20:       omp_offload.cont31:
12166 // CHECK20-NEXT:    [[TMP159:%.*]] = load i32, i32* [[NN]], align 4
12167 // CHECK20-NEXT:    store i32 [[TMP159]], i32* [[NN_CASTED32]], align 4
12168 // CHECK20-NEXT:    [[TMP160:%.*]] = load i32, i32* [[NN_CASTED32]], align 4
12169 // CHECK20-NEXT:    [[TMP161:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
12170 // CHECK20-NEXT:    [[TMP162:%.*]] = bitcast i8** [[TMP161]] to i32*
12171 // CHECK20-NEXT:    store i32 [[TMP160]], i32* [[TMP162]], align 4
12172 // CHECK20-NEXT:    [[TMP163:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
12173 // CHECK20-NEXT:    [[TMP164:%.*]] = bitcast i8** [[TMP163]] to i32*
12174 // CHECK20-NEXT:    store i32 [[TMP160]], i32* [[TMP164]], align 4
12175 // CHECK20-NEXT:    [[TMP165:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS35]], i32 0, i32 0
12176 // CHECK20-NEXT:    store i8* null, i8** [[TMP165]], align 4
12177 // CHECK20-NEXT:    [[TMP166:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS33]], i32 0, i32 0
12178 // CHECK20-NEXT:    [[TMP167:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS34]], i32 0, i32 0
12179 // CHECK20-NEXT:    [[TMP168:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157.region_id, i32 1, i8** [[TMP166]], i8** [[TMP167]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.18, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.19, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12180 // CHECK20-NEXT:    [[TMP169:%.*]] = icmp ne i32 [[TMP168]], 0
12181 // CHECK20-NEXT:    br i1 [[TMP169]], label [[OMP_OFFLOAD_FAILED36:%.*]], label [[OMP_OFFLOAD_CONT37:%.*]]
12182 // CHECK20:       omp_offload.failed36:
12183 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157(i32 [[TMP160]]) #[[ATTR3]]
12184 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT37]]
12185 // CHECK20:       omp_offload.cont37:
12186 // CHECK20-NEXT:    [[TMP170:%.*]] = load i32, i32* [[A]], align 4
12187 // CHECK20-NEXT:    [[TMP171:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
12188 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP171]])
12189 // CHECK20-NEXT:    ret i32 [[TMP170]]
12190 //
12191 //
12192 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
12193 // CHECK20-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR2:[0-9]+]] {
12194 // CHECK20-NEXT:  entry:
12195 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12196 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
12197 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
12198 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12199 // CHECK20-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1]])
12200 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12201 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
12202 // CHECK20-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
12203 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12204 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
12205 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
12206 // CHECK20-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
12207 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
12208 // CHECK20-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12209 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
12210 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12211 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
12212 // CHECK20-NEXT:    ret void
12213 //
12214 //
12215 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
12216 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
12217 // CHECK20-NEXT:  entry:
12218 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12219 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12220 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12221 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12222 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12223 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12224 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12225 // CHECK20-NEXT:    ret void
12226 //
12227 //
12228 // CHECK20-LABEL: define {{[^@]+}}@.omp_task_privates_map.
12229 // 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]+]] {
12230 // CHECK20-NEXT:  entry:
12231 // CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 4
12232 // CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca i16**, align 4
12233 // CHECK20-NEXT:    [[DOTADDR2:%.*]] = alloca [3 x i8*]**, align 4
12234 // CHECK20-NEXT:    [[DOTADDR3:%.*]] = alloca [3 x i8*]**, align 4
12235 // CHECK20-NEXT:    [[DOTADDR4:%.*]] = alloca [3 x i64]**, align 4
12236 // CHECK20-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 4
12237 // CHECK20-NEXT:    store i16** [[TMP1]], i16*** [[DOTADDR1]], align 4
12238 // CHECK20-NEXT:    store [3 x i8*]** [[TMP2]], [3 x i8*]*** [[DOTADDR2]], align 4
12239 // CHECK20-NEXT:    store [3 x i8*]** [[TMP3]], [3 x i8*]*** [[DOTADDR3]], align 4
12240 // CHECK20-NEXT:    store [3 x i64]** [[TMP4]], [3 x i64]*** [[DOTADDR4]], align 4
12241 // CHECK20-NEXT:    [[TMP5:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 4
12242 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 0
12243 // CHECK20-NEXT:    [[TMP7:%.*]] = load [3 x i64]**, [3 x i64]*** [[DOTADDR4]], align 4
12244 // CHECK20-NEXT:    store [3 x i64]* [[TMP6]], [3 x i64]** [[TMP7]], align 4
12245 // CHECK20-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 1
12246 // CHECK20-NEXT:    [[TMP9:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR2]], align 4
12247 // CHECK20-NEXT:    store [3 x i8*]* [[TMP8]], [3 x i8*]** [[TMP9]], align 4
12248 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 2
12249 // CHECK20-NEXT:    [[TMP11:%.*]] = load [3 x i8*]**, [3 x i8*]*** [[DOTADDR3]], align 4
12250 // CHECK20-NEXT:    store [3 x i8*]* [[TMP10]], [3 x i8*]** [[TMP11]], align 4
12251 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T]], %struct..kmp_privates.t* [[TMP5]], i32 0, i32 3
12252 // CHECK20-NEXT:    [[TMP13:%.*]] = load i16**, i16*** [[DOTADDR1]], align 4
12253 // CHECK20-NEXT:    store i16* [[TMP12]], i16** [[TMP13]], align 4
12254 // CHECK20-NEXT:    ret void
12255 //
12256 //
12257 // CHECK20-LABEL: define {{[^@]+}}@.omp_task_entry.
12258 // CHECK20-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
12259 // CHECK20-NEXT:  entry:
12260 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
12261 // CHECK20-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
12262 // CHECK20-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
12263 // CHECK20-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
12264 // CHECK20-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
12265 // CHECK20-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
12266 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i16*, align 4
12267 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR1_I:%.*]] = alloca [3 x i8*]*, align 4
12268 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR2_I:%.*]] = alloca [3 x i8*]*, align 4
12269 // CHECK20-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR3_I:%.*]] = alloca [3 x i64]*, align 4
12270 // CHECK20-NEXT:    [[AA_CASTED_I:%.*]] = alloca i32, align 4
12271 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED_I:%.*]] = alloca i32, align 4
12272 // CHECK20-NEXT:    [[DOTCAPTURE_EXPR__CASTED4_I:%.*]] = alloca i32, align 4
12273 // CHECK20-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
12274 // CHECK20-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
12275 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
12276 // CHECK20-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
12277 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
12278 // CHECK20-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
12279 // CHECK20-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
12280 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
12281 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
12282 // CHECK20-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
12283 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
12284 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
12285 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
12286 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
12287 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
12288 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
12289 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
12290 // CHECK20-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
12291 // CHECK20-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
12292 // CHECK20-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !25
12293 // CHECK20-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
12294 // 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 !25
12295 // CHECK20-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !25
12296 // CHECK20-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
12297 // CHECK20-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !25
12298 // CHECK20-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !25
12299 // CHECK20-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !25
12300 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i16**, [3 x i8*]**, [3 x i8*]**, [3 x i64]**)*
12301 // 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]]
12302 // CHECK20-NEXT:    [[TMP16:%.*]] = load i16*, i16** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 4, !noalias !25
12303 // CHECK20-NEXT:    [[TMP17:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR1_I]], align 4, !noalias !25
12304 // CHECK20-NEXT:    [[TMP18:%.*]] = load [3 x i8*]*, [3 x i8*]** [[DOTFIRSTPRIV_PTR_ADDR2_I]], align 4, !noalias !25
12305 // CHECK20-NEXT:    [[TMP19:%.*]] = load [3 x i64]*, [3 x i64]** [[DOTFIRSTPRIV_PTR_ADDR3_I]], align 4, !noalias !25
12306 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP17]], i32 0, i32 0
12307 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP18]], i32 0, i32 0
12308 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[TMP19]], i32 0, i32 0
12309 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
12310 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
12311 // CHECK20-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP23]], align 4
12312 // CHECK20-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP24]], align 4
12313 // CHECK20-NEXT:    [[TMP27:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101.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]]
12314 // CHECK20-NEXT:    [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
12315 // CHECK20-NEXT:    br i1 [[TMP28]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
12316 // CHECK20:       omp_offload.failed.i:
12317 // CHECK20-NEXT:    [[TMP29:%.*]] = load i16, i16* [[TMP16]], align 2
12318 // CHECK20-NEXT:    [[CONV_I:%.*]] = bitcast i32* [[AA_CASTED_I]] to i16*
12319 // CHECK20-NEXT:    store i16 [[TMP29]], i16* [[CONV_I]], align 2, !noalias !25
12320 // CHECK20-NEXT:    [[TMP30:%.*]] = load i32, i32* [[AA_CASTED_I]], align 4, !noalias !25
12321 // CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[TMP23]], align 4
12322 // CHECK20-NEXT:    store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
12323 // CHECK20-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED_I]], align 4, !noalias !25
12324 // CHECK20-NEXT:    [[TMP33:%.*]] = load i32, i32* [[TMP24]], align 4
12325 // CHECK20-NEXT:    store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
12326 // CHECK20-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED4_I]], align 4, !noalias !25
12327 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP30]], i32 [[TMP32]], i32 [[TMP34]]) #[[ATTR3]]
12328 // CHECK20-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
12329 // CHECK20:       .omp_outlined..1.exit:
12330 // CHECK20-NEXT:    ret i32 0
12331 //
12332 //
12333 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l105
12334 // CHECK20-SAME: (i32 noundef [[A:%.*]]) #[[ATTR2]] {
12335 // CHECK20-NEXT:  entry:
12336 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12337 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12338 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12339 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
12340 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
12341 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
12342 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]])
12343 // CHECK20-NEXT:    ret void
12344 //
12345 //
12346 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
12347 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
12348 // CHECK20-NEXT:  entry:
12349 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12350 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12351 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12352 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12353 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12354 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12355 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
12356 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
12357 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
12358 // CHECK20-NEXT:    ret void
12359 //
12360 //
12361 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
12362 // CHECK20-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR2]] {
12363 // CHECK20-NEXT:  entry:
12364 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12365 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12366 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12367 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12368 // CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
12369 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12370 // CHECK20-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
12371 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12372 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]])
12373 // CHECK20-NEXT:    ret void
12374 //
12375 //
12376 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
12377 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
12378 // CHECK20-NEXT:  entry:
12379 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12380 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12381 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12382 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12383 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12384 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12385 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12386 // CHECK20-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
12387 // CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
12388 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
12389 // CHECK20-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
12390 // CHECK20-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
12391 // CHECK20-NEXT:    ret void
12392 //
12393 //
12394 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
12395 // CHECK20-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
12396 // CHECK20-NEXT:  entry:
12397 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12398 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12399 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12400 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12401 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12402 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12403 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12404 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
12405 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
12406 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
12407 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
12408 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12409 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
12410 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12411 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
12412 // CHECK20-NEXT:    ret void
12413 //
12414 //
12415 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..6
12416 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
12417 // CHECK20-NEXT:  entry:
12418 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12419 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12420 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12421 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12422 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12423 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12424 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12425 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12426 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12427 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
12428 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
12429 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
12430 // CHECK20-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
12431 // CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
12432 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
12433 // CHECK20-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
12434 // CHECK20-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
12435 // CHECK20-NEXT:    ret void
12436 //
12437 //
12438 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
12439 // 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:%.*]]) #[[ATTR2]] {
12440 // CHECK20-NEXT:  entry:
12441 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12442 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
12443 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12444 // CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
12445 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
12446 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
12447 // CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
12448 // CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
12449 // CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
12450 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12451 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12452 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
12453 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12454 // CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
12455 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
12456 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
12457 // CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
12458 // CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
12459 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
12460 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
12461 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12462 // CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
12463 // CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
12464 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
12465 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
12466 // CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
12467 // CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
12468 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
12469 // CHECK20-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
12470 // CHECK20-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
12471 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
12472 // CHECK20-NEXT:    ret void
12473 //
12474 //
12475 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
12476 // 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:%.*]]) #[[ATTR2]] {
12477 // CHECK20-NEXT:  entry:
12478 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12479 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12480 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12481 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
12482 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12483 // CHECK20-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
12484 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
12485 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
12486 // CHECK20-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
12487 // CHECK20-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
12488 // CHECK20-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
12489 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12490 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12491 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12492 // CHECK20-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
12493 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12494 // CHECK20-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
12495 // CHECK20-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
12496 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
12497 // CHECK20-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
12498 // CHECK20-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
12499 // CHECK20-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
12500 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
12501 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12502 // CHECK20-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
12503 // CHECK20-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
12504 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
12505 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
12506 // CHECK20-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
12507 // CHECK20-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
12508 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
12509 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
12510 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
12511 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
12512 // CHECK20-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
12513 // CHECK20-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
12514 // CHECK20-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
12515 // CHECK20-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
12516 // CHECK20-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
12517 // CHECK20-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
12518 // CHECK20-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
12519 // CHECK20-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
12520 // CHECK20-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
12521 // CHECK20-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
12522 // CHECK20-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
12523 // CHECK20-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
12524 // CHECK20-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
12525 // CHECK20-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
12526 // CHECK20-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
12527 // CHECK20-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
12528 // CHECK20-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
12529 // CHECK20-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
12530 // CHECK20-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
12531 // CHECK20-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
12532 // CHECK20-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
12533 // CHECK20-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
12534 // CHECK20-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
12535 // CHECK20-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
12536 // CHECK20-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
12537 // CHECK20-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
12538 // CHECK20-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
12539 // CHECK20-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
12540 // CHECK20-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
12541 // CHECK20-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
12542 // CHECK20-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
12543 // CHECK20-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
12544 // CHECK20-NEXT:    ret void
12545 //
12546 //
12547 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
12548 // CHECK20-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
12549 // CHECK20-NEXT:  entry:
12550 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
12551 // CHECK20-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
12552 // CHECK20-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
12553 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
12554 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
12555 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
12556 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP1]])
12557 // CHECK20-NEXT:    ret void
12558 //
12559 //
12560 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..12
12561 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
12562 // CHECK20-NEXT:  entry:
12563 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12564 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12565 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
12566 // CHECK20-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
12567 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12568 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12569 // CHECK20-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
12570 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
12571 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
12572 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
12573 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP1]])
12574 // CHECK20-NEXT:    ret void
12575 //
12576 //
12577 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..13
12578 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
12579 // CHECK20-NEXT:  entry:
12580 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12581 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12582 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
12583 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12584 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12585 // CHECK20-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
12586 // CHECK20-NEXT:    ret void
12587 //
12588 //
12589 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
12590 // CHECK20-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR2]] {
12591 // CHECK20-NEXT:  entry:
12592 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
12593 // CHECK20-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
12594 // CHECK20-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
12595 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
12596 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
12597 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
12598 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP1]])
12599 // CHECK20-NEXT:    ret void
12600 //
12601 //
12602 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..16
12603 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR2]] {
12604 // CHECK20-NEXT:  entry:
12605 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12606 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12607 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
12608 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12609 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12610 // CHECK20-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
12611 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
12612 // CHECK20-NEXT:    ret void
12613 //
12614 //
12615 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..17
12616 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR2]] {
12617 // CHECK20-NEXT:  entry:
12618 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12619 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12620 // CHECK20-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
12621 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12622 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12623 // CHECK20-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
12624 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
12625 // CHECK20-NEXT:    ret void
12626 //
12627 //
12628 // CHECK20-LABEL: define {{[^@]+}}@_Z6bazzzziPi
12629 // CHECK20-SAME: (i32 noundef [[N:%.*]], i32* noundef [[F:%.*]]) #[[ATTR0]] {
12630 // CHECK20-NEXT:  entry:
12631 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12632 // CHECK20-NEXT:    [[F_ADDR:%.*]] = alloca i32*, align 4
12633 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
12634 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
12635 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
12636 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12637 // CHECK20-NEXT:    store i32* [[F]], i32** [[F_ADDR]], align 4
12638 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
12639 // CHECK20-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12640 // CHECK20-NEXT:    [[TMP2:%.*]] = bitcast i8** [[TMP1]] to i32*
12641 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP2]], align 4
12642 // CHECK20-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12643 // CHECK20-NEXT:    [[TMP4:%.*]] = bitcast i8** [[TMP3]] to i32*
12644 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[TMP4]], align 4
12645 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
12646 // CHECK20-NEXT:    store i8* null, i8** [[TMP5]], align 4
12647 // CHECK20-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12648 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12649 // CHECK20-NEXT:    [[TMP8:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182.region_id, i32 1, i8** [[TMP6]], i8** [[TMP7]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.21, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.22, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12650 // CHECK20-NEXT:    [[TMP9:%.*]] = icmp ne i32 [[TMP8]], 0
12651 // CHECK20-NEXT:    br i1 [[TMP9]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12652 // CHECK20:       omp_offload.failed:
12653 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182(i32 [[TMP0]]) #[[ATTR3]]
12654 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12655 // CHECK20:       omp_offload.cont:
12656 // CHECK20-NEXT:    ret void
12657 //
12658 //
12659 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
12660 // CHECK20-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
12661 // CHECK20-NEXT:  entry:
12662 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12663 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12664 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12665 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..20 to void (i32*, i32*, ...)*), i32 [[TMP0]])
12666 // CHECK20-NEXT:    ret void
12667 //
12668 //
12669 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..20
12670 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR2]] {
12671 // CHECK20-NEXT:  entry:
12672 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12673 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12674 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12675 // CHECK20-NEXT:    [[F:%.*]] = alloca i32*, align 4
12676 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12677 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12678 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12679 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12680 // CHECK20-NEXT:    ret void
12681 //
12682 //
12683 // CHECK20-LABEL: define {{[^@]+}}@_Z3bari
12684 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
12685 // CHECK20-NEXT:  entry:
12686 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12687 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
12688 // CHECK20-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
12689 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12690 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
12691 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
12692 // CHECK20-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
12693 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
12694 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
12695 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
12696 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
12697 // CHECK20-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
12698 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
12699 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
12700 // CHECK20-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
12701 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
12702 // CHECK20-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
12703 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
12704 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
12705 // CHECK20-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
12706 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
12707 // CHECK20-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
12708 // CHECK20-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
12709 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
12710 // CHECK20-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
12711 // CHECK20-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
12712 // CHECK20-NEXT:    ret i32 [[TMP8]]
12713 //
12714 //
12715 // CHECK20-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
12716 // CHECK20-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
12717 // CHECK20-NEXT:  entry:
12718 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
12719 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12720 // CHECK20-NEXT:    [[B:%.*]] = alloca i32, align 4
12721 // CHECK20-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
12722 // CHECK20-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
12723 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
12724 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
12725 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
12726 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
12727 // CHECK20-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
12728 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
12729 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12730 // CHECK20-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
12731 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
12732 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
12733 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
12734 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
12735 // CHECK20-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
12736 // CHECK20-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
12737 // CHECK20-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
12738 // CHECK20-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
12739 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
12740 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
12741 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
12742 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
12743 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
12744 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
12745 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12746 // CHECK20:       omp_if.then:
12747 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
12748 // CHECK20-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
12749 // CHECK20-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
12750 // CHECK20-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
12751 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
12752 // CHECK20-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.24 to i8*), i32 40, i1 false)
12753 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12754 // CHECK20-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
12755 // CHECK20-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
12756 // CHECK20-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12757 // CHECK20-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
12758 // CHECK20-NEXT:    store double* [[A]], double** [[TMP14]], align 4
12759 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
12760 // CHECK20-NEXT:    store i8* null, i8** [[TMP15]], align 4
12761 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
12762 // CHECK20-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
12763 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
12764 // CHECK20-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
12765 // CHECK20-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
12766 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
12767 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
12768 // CHECK20-NEXT:    store i8* null, i8** [[TMP20]], align 4
12769 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
12770 // CHECK20-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
12771 // CHECK20-NEXT:    store i32 2, i32* [[TMP22]], align 4
12772 // CHECK20-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
12773 // CHECK20-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
12774 // CHECK20-NEXT:    store i32 2, i32* [[TMP24]], align 4
12775 // CHECK20-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
12776 // CHECK20-NEXT:    store i8* null, i8** [[TMP25]], align 4
12777 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
12778 // CHECK20-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
12779 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
12780 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
12781 // CHECK20-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
12782 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
12783 // CHECK20-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
12784 // CHECK20-NEXT:    store i8* null, i8** [[TMP30]], align 4
12785 // CHECK20-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
12786 // CHECK20-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
12787 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
12788 // CHECK20-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
12789 // CHECK20-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
12790 // CHECK20-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
12791 // CHECK20-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
12792 // CHECK20-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
12793 // CHECK20-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
12794 // CHECK20-NEXT:    store i8* null, i8** [[TMP36]], align 4
12795 // CHECK20-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12796 // CHECK20-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12797 // CHECK20-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
12798 // CHECK20-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.25, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12799 // CHECK20-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
12800 // CHECK20-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12801 // CHECK20:       omp_offload.failed:
12802 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
12803 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12804 // CHECK20:       omp_offload.cont:
12805 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
12806 // CHECK20:       omp_if.else:
12807 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR3]]
12808 // CHECK20-NEXT:    br label [[OMP_IF_END]]
12809 // CHECK20:       omp_if.end:
12810 // CHECK20-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
12811 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
12812 // CHECK20-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
12813 // CHECK20-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
12814 // CHECK20-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
12815 // CHECK20-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
12816 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
12817 // CHECK20-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
12818 // CHECK20-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
12819 // CHECK20-NEXT:    ret i32 [[ADD3]]
12820 //
12821 //
12822 // CHECK20-LABEL: define {{[^@]+}}@_ZL7fstatici
12823 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
12824 // CHECK20-NEXT:  entry:
12825 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12826 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
12827 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
12828 // CHECK20-NEXT:    [[AAA:%.*]] = alloca i8, align 1
12829 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
12830 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12831 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12832 // CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
12833 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
12834 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
12835 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
12836 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12837 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
12838 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
12839 // CHECK20-NEXT:    store i8 0, i8* [[AAA]], align 1
12840 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
12841 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
12842 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
12843 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
12844 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12845 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
12846 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12847 // CHECK20-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
12848 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
12849 // CHECK20-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
12850 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
12851 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
12852 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
12853 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12854 // CHECK20:       omp_if.then:
12855 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12856 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
12857 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
12858 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12859 // CHECK20-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
12860 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
12861 // CHECK20-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
12862 // CHECK20-NEXT:    store i8* null, i8** [[TMP11]], align 4
12863 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
12864 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
12865 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
12866 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
12867 // CHECK20-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
12868 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
12869 // CHECK20-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
12870 // CHECK20-NEXT:    store i8* null, i8** [[TMP16]], align 4
12871 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
12872 // CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
12873 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
12874 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
12875 // CHECK20-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
12876 // CHECK20-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
12877 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
12878 // CHECK20-NEXT:    store i8* null, i8** [[TMP21]], align 4
12879 // CHECK20-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
12880 // CHECK20-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
12881 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
12882 // CHECK20-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
12883 // CHECK20-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
12884 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
12885 // CHECK20-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
12886 // CHECK20-NEXT:    store i8* null, i8** [[TMP26]], align 4
12887 // CHECK20-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12888 // CHECK20-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12889 // CHECK20-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12890 // CHECK20-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
12891 // CHECK20-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12892 // CHECK20:       omp_offload.failed:
12893 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
12894 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12895 // CHECK20:       omp_offload.cont:
12896 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
12897 // CHECK20:       omp_if.else:
12898 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR3]]
12899 // CHECK20-NEXT:    br label [[OMP_IF_END]]
12900 // CHECK20:       omp_if.end:
12901 // CHECK20-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
12902 // CHECK20-NEXT:    ret i32 [[TMP31]]
12903 //
12904 //
12905 // CHECK20-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
12906 // CHECK20-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
12907 // CHECK20-NEXT:  entry:
12908 // CHECK20-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
12909 // CHECK20-NEXT:    [[A:%.*]] = alloca i32, align 4
12910 // CHECK20-NEXT:    [[AA:%.*]] = alloca i16, align 2
12911 // CHECK20-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
12912 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12913 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12914 // CHECK20-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
12915 // CHECK20-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
12916 // CHECK20-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
12917 // CHECK20-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
12918 // CHECK20-NEXT:    store i32 0, i32* [[A]], align 4
12919 // CHECK20-NEXT:    store i16 0, i16* [[AA]], align 2
12920 // CHECK20-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
12921 // CHECK20-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
12922 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
12923 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
12924 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12925 // CHECK20-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
12926 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12927 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
12928 // CHECK20-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
12929 // CHECK20-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12930 // CHECK20:       omp_if.then:
12931 // CHECK20-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12932 // CHECK20-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
12933 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
12934 // CHECK20-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12935 // CHECK20-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
12936 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
12937 // CHECK20-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
12938 // CHECK20-NEXT:    store i8* null, i8** [[TMP9]], align 4
12939 // CHECK20-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
12940 // CHECK20-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
12941 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
12942 // CHECK20-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
12943 // CHECK20-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
12944 // CHECK20-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
12945 // CHECK20-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
12946 // CHECK20-NEXT:    store i8* null, i8** [[TMP14]], align 4
12947 // CHECK20-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
12948 // CHECK20-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
12949 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
12950 // CHECK20-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
12951 // CHECK20-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
12952 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
12953 // CHECK20-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
12954 // CHECK20-NEXT:    store i8* null, i8** [[TMP19]], align 4
12955 // CHECK20-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12956 // CHECK20-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12957 // CHECK20-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB1]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.30, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.31, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12958 // CHECK20-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
12959 // CHECK20-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12960 // CHECK20:       omp_offload.failed:
12961 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
12962 // CHECK20-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12963 // CHECK20:       omp_offload.cont:
12964 // CHECK20-NEXT:    br label [[OMP_IF_END:%.*]]
12965 // CHECK20:       omp_if.else:
12966 // CHECK20-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR3]]
12967 // CHECK20-NEXT:    br label [[OMP_IF_END]]
12968 // CHECK20:       omp_if.end:
12969 // CHECK20-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
12970 // CHECK20-NEXT:    ret i32 [[TMP24]]
12971 //
12972 //
12973 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
12974 // CHECK20-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
12975 // CHECK20-NEXT:  entry:
12976 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
12977 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
12978 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
12979 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
12980 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
12981 // CHECK20-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
12982 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
12983 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
12984 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12985 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
12986 // CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
12987 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
12988 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12989 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
12990 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
12991 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
12992 // CHECK20-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
12993 // CHECK20-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
12994 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..23 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
12995 // CHECK20-NEXT:    ret void
12996 //
12997 //
12998 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..23
12999 // 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]] {
13000 // CHECK20-NEXT:  entry:
13001 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13002 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13003 // CHECK20-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
13004 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
13005 // CHECK20-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
13006 // CHECK20-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
13007 // CHECK20-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
13008 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13009 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13010 // CHECK20-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
13011 // CHECK20-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
13012 // CHECK20-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13013 // CHECK20-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
13014 // CHECK20-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
13015 // CHECK20-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
13016 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13017 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
13018 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
13019 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
13020 // CHECK20-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
13021 // CHECK20-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
13022 // CHECK20-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
13023 // CHECK20-NEXT:    store double [[ADD]], double* [[A]], align 4
13024 // CHECK20-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
13025 // CHECK20-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
13026 // CHECK20-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
13027 // CHECK20-NEXT:    store double [[INC]], double* [[A3]], align 4
13028 // CHECK20-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
13029 // CHECK20-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
13030 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
13031 // CHECK20-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
13032 // CHECK20-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
13033 // CHECK20-NEXT:    ret void
13034 //
13035 //
13036 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
13037 // CHECK20-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13038 // CHECK20-NEXT:  entry:
13039 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
13040 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
13041 // CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
13042 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
13043 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
13044 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
13045 // CHECK20-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
13046 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
13047 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
13048 // CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
13049 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
13050 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
13051 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
13052 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
13053 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
13054 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
13055 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
13056 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
13057 // CHECK20-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
13058 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
13059 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
13060 // CHECK20-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
13061 // CHECK20-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
13062 // CHECK20-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
13063 // CHECK20-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
13064 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
13065 // CHECK20-NEXT:    ret void
13066 //
13067 //
13068 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..26
13069 // CHECK20-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13070 // CHECK20-NEXT:  entry:
13071 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13072 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13073 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
13074 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
13075 // CHECK20-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
13076 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
13077 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13078 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13079 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
13080 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
13081 // CHECK20-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
13082 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
13083 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
13084 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
13085 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
13086 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
13087 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
13088 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
13089 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
13090 // CHECK20-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
13091 // CHECK20-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
13092 // CHECK20-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
13093 // CHECK20-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
13094 // CHECK20-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
13095 // CHECK20-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
13096 // CHECK20-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
13097 // CHECK20-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
13098 // CHECK20-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
13099 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
13100 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13101 // CHECK20-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
13102 // CHECK20-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
13103 // CHECK20-NEXT:    ret void
13104 //
13105 //
13106 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
13107 // CHECK20-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
13108 // CHECK20-NEXT:  entry:
13109 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
13110 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
13111 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
13112 // CHECK20-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
13113 // CHECK20-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
13114 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
13115 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
13116 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
13117 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
13118 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
13119 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
13120 // CHECK20-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
13121 // CHECK20-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
13122 // CHECK20-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
13123 // CHECK20-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
13124 // CHECK20-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
13125 // CHECK20-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
13126 // CHECK20-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..29 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
13127 // CHECK20-NEXT:    ret void
13128 //
13129 //
13130 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..29
13131 // 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]] {
13132 // CHECK20-NEXT:  entry:
13133 // CHECK20-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13134 // CHECK20-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13135 // CHECK20-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
13136 // CHECK20-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
13137 // CHECK20-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
13138 // CHECK20-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13139 // CHECK20-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13140 // CHECK20-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
13141 // CHECK20-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
13142 // CHECK20-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
13143 // CHECK20-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
13144 // CHECK20-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
13145 // CHECK20-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
13146 // CHECK20-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
13147 // CHECK20-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
13148 // CHECK20-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
13149 // CHECK20-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
13150 // CHECK20-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
13151 // CHECK20-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
13152 // CHECK20-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
13153 // CHECK20-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
13154 // CHECK20-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13155 // CHECK20-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
13156 // CHECK20-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
13157 // CHECK20-NEXT:    ret void
13158 //
13159 //
13160 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
13161 // CHECK20-SAME: () #[[ATTR4]] {
13162 // CHECK20-NEXT:  entry:
13163 // CHECK20-NEXT:    call void @__tgt_register_requires(i64 1)
13164 // CHECK20-NEXT:    ret void
13165 //
13166 //
13167 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
13168 // CHECK25-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
13169 // CHECK25-NEXT:  entry:
13170 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13171 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
13172 // CHECK25-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
13173 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13174 // CHECK25-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
13175 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13176 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
13177 // CHECK25-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
13178 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13179 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
13180 // CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
13181 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
13182 // CHECK25-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
13183 // CHECK25-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
13184 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
13185 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13186 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
13187 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13188 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
13189 // CHECK25-NEXT:    ret void
13190 //
13191 //
13192 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
13193 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13194 // CHECK25-NEXT:  entry:
13195 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13196 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13197 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13198 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13199 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13200 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13201 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13202 // CHECK25-NEXT:    ret void
13203 //
13204 //
13205 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
13206 // CHECK25-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13207 // CHECK25-NEXT:  entry:
13208 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13209 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13210 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13211 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13212 // CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
13213 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13214 // CHECK25-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
13215 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13216 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13217 // CHECK25-NEXT:    ret void
13218 //
13219 //
13220 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
13221 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13222 // CHECK25-NEXT:  entry:
13223 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13224 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13225 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13226 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13227 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13228 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13229 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13230 // CHECK25-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
13231 // CHECK25-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
13232 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
13233 // CHECK25-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
13234 // CHECK25-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
13235 // CHECK25-NEXT:    ret void
13236 //
13237 //
13238 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
13239 // CHECK25-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13240 // CHECK25-NEXT:  entry:
13241 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13242 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13243 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13244 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13245 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13246 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13247 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13248 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13249 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13250 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13251 // CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
13252 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
13253 // CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
13254 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13255 // CHECK25-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
13256 // CHECK25-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13257 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
13258 // CHECK25-NEXT:    ret void
13259 //
13260 //
13261 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..2
13262 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13263 // CHECK25-NEXT:  entry:
13264 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13265 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13266 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13267 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13268 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13269 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13270 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13271 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13272 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13273 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13274 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13275 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
13276 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13277 // CHECK25-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
13278 // CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
13279 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
13280 // CHECK25-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
13281 // CHECK25-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
13282 // CHECK25-NEXT:    ret void
13283 //
13284 //
13285 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
13286 // 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:%.*]]) #[[ATTR0]] {
13287 // CHECK25-NEXT:  entry:
13288 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13289 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
13290 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13291 // CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
13292 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
13293 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13294 // CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
13295 // CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
13296 // CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
13297 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13298 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13299 // CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
13300 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13301 // CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
13302 // CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
13303 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13304 // CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
13305 // CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
13306 // CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
13307 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13308 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
13309 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13310 // CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
13311 // CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
13312 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13313 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
13314 // CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
13315 // CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
13316 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
13317 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13318 // CHECK25-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
13319 // CHECK25-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
13320 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
13321 // CHECK25-NEXT:    ret void
13322 //
13323 //
13324 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..3
13325 // 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:%.*]]) #[[ATTR0]] {
13326 // CHECK25-NEXT:  entry:
13327 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13328 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13329 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13330 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
13331 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13332 // CHECK25-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
13333 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
13334 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13335 // CHECK25-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
13336 // CHECK25-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
13337 // CHECK25-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
13338 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13339 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13340 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13341 // CHECK25-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
13342 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13343 // CHECK25-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
13344 // CHECK25-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
13345 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13346 // CHECK25-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
13347 // CHECK25-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
13348 // CHECK25-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
13349 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13350 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
13351 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13352 // CHECK25-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
13353 // CHECK25-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
13354 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13355 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
13356 // CHECK25-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
13357 // CHECK25-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
13358 // CHECK25-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
13359 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
13360 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13361 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
13362 // CHECK25-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
13363 // CHECK25-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
13364 // CHECK25-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
13365 // CHECK25-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
13366 // CHECK25-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
13367 // CHECK25-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
13368 // CHECK25-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
13369 // CHECK25-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
13370 // CHECK25-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
13371 // CHECK25-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
13372 // CHECK25-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
13373 // CHECK25-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
13374 // CHECK25-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
13375 // CHECK25-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
13376 // CHECK25-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
13377 // CHECK25-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
13378 // CHECK25-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
13379 // CHECK25-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
13380 // CHECK25-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
13381 // CHECK25-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
13382 // CHECK25-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
13383 // CHECK25-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
13384 // CHECK25-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
13385 // CHECK25-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
13386 // CHECK25-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
13387 // CHECK25-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
13388 // CHECK25-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
13389 // CHECK25-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
13390 // CHECK25-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
13391 // CHECK25-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
13392 // CHECK25-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
13393 // CHECK25-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
13394 // CHECK25-NEXT:    ret void
13395 //
13396 //
13397 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
13398 // CHECK25-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13399 // CHECK25-NEXT:  entry:
13400 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13401 // CHECK25-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13402 // CHECK25-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13403 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13404 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13405 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13406 // CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13407 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13408 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13409 // CHECK25-NEXT:    ret void
13410 //
13411 //
13412 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..4
13413 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13414 // CHECK25-NEXT:  entry:
13415 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13416 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13417 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13418 // CHECK25-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13419 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13420 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13421 // CHECK25-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13422 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13423 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13424 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13425 // CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13426 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13427 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13428 // CHECK25-NEXT:    ret void
13429 //
13430 //
13431 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..5
13432 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13433 // CHECK25-NEXT:  entry:
13434 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13435 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13436 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13437 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13438 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13439 // CHECK25-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13440 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13441 // CHECK25-NEXT:    ret void
13442 //
13443 //
13444 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
13445 // CHECK25-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13446 // CHECK25-NEXT:  entry:
13447 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13448 // CHECK25-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13449 // CHECK25-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13450 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13451 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13452 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13453 // CHECK25-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13454 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13455 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13456 // CHECK25-NEXT:    ret void
13457 //
13458 //
13459 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..6
13460 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13461 // CHECK25-NEXT:  entry:
13462 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13463 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13464 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13465 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13466 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13467 // CHECK25-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13468 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13469 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[CONV]])
13470 // CHECK25-NEXT:    ret void
13471 //
13472 //
13473 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..7
13474 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
13475 // CHECK25-NEXT:  entry:
13476 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13477 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13478 // CHECK25-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
13479 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13480 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13481 // CHECK25-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
13482 // CHECK25-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
13483 // CHECK25-NEXT:    ret void
13484 //
13485 //
13486 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
13487 // CHECK25-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
13488 // CHECK25-NEXT:  entry:
13489 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13490 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13491 // CHECK25-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13492 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i64 [[TMP0]])
13493 // CHECK25-NEXT:    ret void
13494 //
13495 //
13496 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..8
13497 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
13498 // CHECK25-NEXT:  entry:
13499 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13500 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13501 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13502 // CHECK25-NEXT:    [[F:%.*]] = alloca i32*, align 8
13503 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13504 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13505 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13506 // CHECK25-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13507 // CHECK25-NEXT:    ret void
13508 //
13509 //
13510 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
13511 // CHECK25-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
13512 // CHECK25-NEXT:  entry:
13513 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13514 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13515 // CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
13516 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13517 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13518 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13519 // CHECK25-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
13520 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13521 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13522 // CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
13523 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13524 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13525 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13526 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
13527 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13528 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13529 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13530 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
13531 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
13532 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
13533 // CHECK25-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13534 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
13535 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13536 // CHECK25-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
13537 // CHECK25-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
13538 // CHECK25-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
13539 // CHECK25-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
13540 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
13541 // CHECK25-NEXT:    ret void
13542 //
13543 //
13544 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..9
13545 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
13546 // CHECK25-NEXT:  entry:
13547 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13548 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13549 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13550 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13551 // CHECK25-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
13552 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13553 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13554 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13555 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13556 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13557 // CHECK25-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
13558 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13559 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13560 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13561 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
13562 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13563 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13564 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
13565 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13566 // CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
13567 // CHECK25-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
13568 // CHECK25-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
13569 // CHECK25-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
13570 // CHECK25-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
13571 // CHECK25-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
13572 // CHECK25-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
13573 // CHECK25-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
13574 // CHECK25-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
13575 // CHECK25-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
13576 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
13577 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13578 // CHECK25-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
13579 // CHECK25-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
13580 // CHECK25-NEXT:    ret void
13581 //
13582 //
13583 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
13584 // CHECK25-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
13585 // CHECK25-NEXT:  entry:
13586 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
13587 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
13588 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13589 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13590 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
13591 // CHECK25-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
13592 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
13593 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
13594 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13595 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13596 // CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
13597 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
13598 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
13599 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13600 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13601 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
13602 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
13603 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
13604 // CHECK25-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
13605 // CHECK25-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
13606 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
13607 // CHECK25-NEXT:    ret void
13608 //
13609 //
13610 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..10
13611 // 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]] {
13612 // CHECK25-NEXT:  entry:
13613 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13614 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13615 // CHECK25-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
13616 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
13617 // CHECK25-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13618 // CHECK25-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13619 // CHECK25-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
13620 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13621 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13622 // CHECK25-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
13623 // CHECK25-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
13624 // CHECK25-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13625 // CHECK25-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13626 // CHECK25-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
13627 // CHECK25-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
13628 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
13629 // CHECK25-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13630 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13631 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
13632 // CHECK25-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
13633 // CHECK25-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
13634 // CHECK25-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
13635 // CHECK25-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
13636 // CHECK25-NEXT:    store double [[ADD]], double* [[A]], align 8
13637 // CHECK25-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
13638 // CHECK25-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
13639 // CHECK25-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
13640 // CHECK25-NEXT:    store double [[INC]], double* [[A4]], align 8
13641 // CHECK25-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
13642 // CHECK25-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
13643 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
13644 // CHECK25-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
13645 // CHECK25-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
13646 // CHECK25-NEXT:    ret void
13647 //
13648 //
13649 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
13650 // CHECK25-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
13651 // CHECK25-NEXT:  entry:
13652 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13653 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13654 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13655 // CHECK25-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13656 // CHECK25-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13657 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13658 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13659 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13660 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13661 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13662 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13663 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13664 // CHECK25-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13665 // CHECK25-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
13666 // CHECK25-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
13667 // CHECK25-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
13668 // CHECK25-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13669 // CHECK25-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
13670 // CHECK25-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13671 // CHECK25-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
13672 // CHECK25-NEXT:    ret void
13673 //
13674 //
13675 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..11
13676 // 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]] {
13677 // CHECK25-NEXT:  entry:
13678 // CHECK25-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13679 // CHECK25-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13680 // CHECK25-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13681 // CHECK25-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13682 // CHECK25-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
13683 // CHECK25-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13684 // CHECK25-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13685 // CHECK25-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13686 // CHECK25-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13687 // CHECK25-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
13688 // CHECK25-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13689 // CHECK25-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13690 // CHECK25-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
13691 // CHECK25-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
13692 // CHECK25-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
13693 // CHECK25-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13694 // CHECK25-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
13695 // CHECK25-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
13696 // CHECK25-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
13697 // CHECK25-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
13698 // CHECK25-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
13699 // CHECK25-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
13700 // CHECK25-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
13701 // CHECK25-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
13702 // CHECK25-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
13703 // CHECK25-NEXT:    ret void
13704 //
13705 //
13706 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
13707 // CHECK26-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
13708 // CHECK26-NEXT:  entry:
13709 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13710 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
13711 // CHECK26-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i64, align 8
13712 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13713 // CHECK26-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
13714 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13715 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
13716 // CHECK26-NEXT:    store i64 [[DOTCAPTURE_EXPR_1]], i64* [[DOTCAPTURE_EXPR__ADDR2]], align 8
13717 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13718 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
13719 // CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR2]] to i32*
13720 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV3]], align 4
13721 // CHECK26-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV4]], align 4
13722 // CHECK26-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
13723 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
13724 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13725 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV5]], align 2
13726 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13727 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]])
13728 // CHECK26-NEXT:    ret void
13729 //
13730 //
13731 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
13732 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13733 // CHECK26-NEXT:  entry:
13734 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13735 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13736 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13737 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13738 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13739 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13740 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13741 // CHECK26-NEXT:    ret void
13742 //
13743 //
13744 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
13745 // CHECK26-SAME: (i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13746 // CHECK26-NEXT:  entry:
13747 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13748 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13749 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13750 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13751 // CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
13752 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13753 // CHECK26-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
13754 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13755 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13756 // CHECK26-NEXT:    ret void
13757 //
13758 //
13759 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
13760 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13761 // CHECK26-NEXT:  entry:
13762 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13763 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13764 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13765 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13766 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13767 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13768 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13769 // CHECK26-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
13770 // CHECK26-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
13771 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
13772 // CHECK26-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
13773 // CHECK26-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
13774 // CHECK26-NEXT:    ret void
13775 //
13776 //
13777 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
13778 // CHECK26-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13779 // CHECK26-NEXT:  entry:
13780 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13781 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13782 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13783 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
13784 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13785 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13786 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13787 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13788 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13789 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13790 // CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
13791 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
13792 // CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
13793 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
13794 // CHECK26-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
13795 // CHECK26-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
13796 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
13797 // CHECK26-NEXT:    ret void
13798 //
13799 //
13800 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..2
13801 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
13802 // CHECK26-NEXT:  entry:
13803 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13804 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13805 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13806 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
13807 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13808 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13809 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13810 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
13811 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13812 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
13813 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13814 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
13815 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13816 // CHECK26-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV1]], align 2
13817 // CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP1]] to i32
13818 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
13819 // CHECK26-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
13820 // CHECK26-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
13821 // CHECK26-NEXT:    ret void
13822 //
13823 //
13824 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
13825 // 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:%.*]]) #[[ATTR0]] {
13826 // CHECK26-NEXT:  entry:
13827 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13828 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
13829 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13830 // CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
13831 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
13832 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13833 // CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
13834 // CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
13835 // CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
13836 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
13837 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13838 // CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
13839 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13840 // CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
13841 // CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
13842 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13843 // CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
13844 // CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
13845 // CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
13846 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13847 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
13848 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13849 // CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
13850 // CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
13851 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13852 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
13853 // CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
13854 // CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
13855 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
13856 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
13857 // CHECK26-NEXT:    store i32 [[TMP8]], i32* [[CONV5]], align 4
13858 // CHECK26-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
13859 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*)* @.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]])
13860 // CHECK26-NEXT:    ret void
13861 //
13862 //
13863 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..3
13864 // 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:%.*]]) #[[ATTR0]] {
13865 // CHECK26-NEXT:  entry:
13866 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13867 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13868 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
13869 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
13870 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
13871 // CHECK26-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
13872 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
13873 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
13874 // CHECK26-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
13875 // CHECK26-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
13876 // CHECK26-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
13877 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13878 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13879 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
13880 // CHECK26-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
13881 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
13882 // CHECK26-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
13883 // CHECK26-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
13884 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
13885 // CHECK26-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
13886 // CHECK26-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
13887 // CHECK26-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
13888 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
13889 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
13890 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
13891 // CHECK26-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
13892 // CHECK26-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
13893 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
13894 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
13895 // CHECK26-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
13896 // CHECK26-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
13897 // CHECK26-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
13898 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
13899 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
13900 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
13901 // CHECK26-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
13902 // CHECK26-NEXT:    [[CONV5:%.*]] = fpext float [[TMP9]] to double
13903 // CHECK26-NEXT:    [[ADD6:%.*]] = fadd double [[CONV5]], 1.000000e+00
13904 // CHECK26-NEXT:    [[CONV7:%.*]] = fptrunc double [[ADD6]] to float
13905 // CHECK26-NEXT:    store float [[CONV7]], float* [[ARRAYIDX]], align 4
13906 // CHECK26-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
13907 // CHECK26-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX8]], align 4
13908 // CHECK26-NEXT:    [[CONV9:%.*]] = fpext float [[TMP10]] to double
13909 // CHECK26-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
13910 // CHECK26-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
13911 // CHECK26-NEXT:    store float [[CONV11]], float* [[ARRAYIDX8]], align 4
13912 // CHECK26-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
13913 // CHECK26-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX12]], i64 0, i64 2
13914 // CHECK26-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX13]], align 8
13915 // CHECK26-NEXT:    [[ADD14:%.*]] = fadd double [[TMP11]], 1.000000e+00
13916 // CHECK26-NEXT:    store double [[ADD14]], double* [[ARRAYIDX13]], align 8
13917 // CHECK26-NEXT:    [[TMP12:%.*]] = mul nsw i64 1, [[TMP5]]
13918 // CHECK26-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP12]]
13919 // CHECK26-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX15]], i64 3
13920 // CHECK26-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX16]], align 8
13921 // CHECK26-NEXT:    [[ADD17:%.*]] = fadd double [[TMP13]], 1.000000e+00
13922 // CHECK26-NEXT:    store double [[ADD17]], double* [[ARRAYIDX16]], align 8
13923 // CHECK26-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
13924 // CHECK26-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 8
13925 // CHECK26-NEXT:    [[ADD18:%.*]] = add nsw i64 [[TMP14]], 1
13926 // CHECK26-NEXT:    store i64 [[ADD18]], i64* [[X]], align 8
13927 // CHECK26-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
13928 // CHECK26-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 8
13929 // CHECK26-NEXT:    [[CONV19:%.*]] = sext i8 [[TMP15]] to i32
13930 // CHECK26-NEXT:    [[ADD20:%.*]] = add nsw i32 [[CONV19]], 1
13931 // CHECK26-NEXT:    [[CONV21:%.*]] = trunc i32 [[ADD20]] to i8
13932 // CHECK26-NEXT:    store i8 [[CONV21]], i8* [[Y]], align 8
13933 // CHECK26-NEXT:    ret void
13934 //
13935 //
13936 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
13937 // CHECK26-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13938 // CHECK26-NEXT:  entry:
13939 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13940 // CHECK26-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13941 // CHECK26-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13942 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13943 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13944 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13945 // CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13946 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13947 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13948 // CHECK26-NEXT:    ret void
13949 //
13950 //
13951 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..4
13952 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13953 // CHECK26-NEXT:  entry:
13954 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13955 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13956 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13957 // CHECK26-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13958 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13959 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13960 // CHECK26-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13961 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13962 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13963 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13964 // CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13965 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13966 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13967 // CHECK26-NEXT:    ret void
13968 //
13969 //
13970 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..5
13971 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13972 // CHECK26-NEXT:  entry:
13973 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13974 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13975 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13976 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13977 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13978 // CHECK26-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13979 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13980 // CHECK26-NEXT:    ret void
13981 //
13982 //
13983 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
13984 // CHECK26-SAME: (i64 noundef [[NN:%.*]]) #[[ATTR0]] {
13985 // CHECK26-NEXT:  entry:
13986 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
13987 // CHECK26-NEXT:    [[NN_CASTED:%.*]] = alloca i64, align 8
13988 // CHECK26-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
13989 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
13990 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
13991 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[NN_CASTED]] to i32*
13992 // CHECK26-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
13993 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[NN_CASTED]], align 8
13994 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP1]])
13995 // CHECK26-NEXT:    ret void
13996 //
13997 //
13998 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..6
13999 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[NN:%.*]]) #[[ATTR0]] {
14000 // CHECK26-NEXT:  entry:
14001 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14002 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14003 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i64, align 8
14004 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14005 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14006 // CHECK26-NEXT:    store i64 [[NN]], i64* [[NN_ADDR]], align 8
14007 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[NN_ADDR]] to i32*
14008 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[CONV]])
14009 // CHECK26-NEXT:    ret void
14010 //
14011 //
14012 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..7
14013 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
14014 // CHECK26-NEXT:  entry:
14015 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14016 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14017 // CHECK26-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 8
14018 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14019 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14020 // CHECK26-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 8
14021 // CHECK26-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 8
14022 // CHECK26-NEXT:    ret void
14023 //
14024 //
14025 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
14026 // CHECK26-SAME: (i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
14027 // CHECK26-NEXT:  entry:
14028 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
14029 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
14030 // CHECK26-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
14031 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i64 [[TMP0]])
14032 // CHECK26-NEXT:    ret void
14033 //
14034 //
14035 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..8
14036 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[VLA:%.*]]) #[[ATTR0]] {
14037 // CHECK26-NEXT:  entry:
14038 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14039 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14040 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
14041 // CHECK26-NEXT:    [[F:%.*]] = alloca i32*, align 8
14042 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14043 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14044 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
14045 // CHECK26-NEXT:    [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
14046 // CHECK26-NEXT:    ret void
14047 //
14048 //
14049 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
14050 // CHECK26-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14051 // CHECK26-NEXT:  entry:
14052 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
14053 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
14054 // CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
14055 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
14056 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
14057 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
14058 // CHECK26-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
14059 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
14060 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
14061 // CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
14062 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
14063 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
14064 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
14065 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
14066 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
14067 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
14068 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
14069 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
14070 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
14071 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
14072 // CHECK26-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
14073 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
14074 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
14075 // CHECK26-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
14076 // CHECK26-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
14077 // CHECK26-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
14078 // CHECK26-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
14079 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
14080 // CHECK26-NEXT:    ret void
14081 //
14082 //
14083 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..9
14084 // CHECK26-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14085 // CHECK26-NEXT:  entry:
14086 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14087 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14088 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
14089 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
14090 // CHECK26-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
14091 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
14092 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14093 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14094 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
14095 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
14096 // CHECK26-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
14097 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
14098 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
14099 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
14100 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
14101 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
14102 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
14103 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
14104 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
14105 // CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
14106 // CHECK26-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP2]] to i32
14107 // CHECK26-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
14108 // CHECK26-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
14109 // CHECK26-NEXT:    store i16 [[CONV5]], i16* [[CONV1]], align 2
14110 // CHECK26-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV2]], align 1
14111 // CHECK26-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP3]] to i32
14112 // CHECK26-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], 1
14113 // CHECK26-NEXT:    [[CONV8:%.*]] = trunc i32 [[ADD7]] to i8
14114 // CHECK26-NEXT:    store i8 [[CONV8]], i8* [[CONV2]], align 1
14115 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
14116 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
14117 // CHECK26-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP4]], 1
14118 // CHECK26-NEXT:    store i32 [[ADD9]], i32* [[ARRAYIDX]], align 4
14119 // CHECK26-NEXT:    ret void
14120 //
14121 //
14122 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
14123 // CHECK26-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
14124 // CHECK26-NEXT:  entry:
14125 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
14126 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
14127 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
14128 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
14129 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
14130 // CHECK26-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
14131 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
14132 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
14133 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
14134 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
14135 // CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
14136 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
14137 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
14138 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
14139 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
14140 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
14141 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
14142 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
14143 // CHECK26-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
14144 // CHECK26-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
14145 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
14146 // CHECK26-NEXT:    ret void
14147 //
14148 //
14149 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..10
14150 // 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]] {
14151 // CHECK26-NEXT:  entry:
14152 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14153 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14154 // CHECK26-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
14155 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
14156 // CHECK26-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
14157 // CHECK26-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
14158 // CHECK26-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
14159 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14160 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14161 // CHECK26-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
14162 // CHECK26-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
14163 // CHECK26-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
14164 // CHECK26-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
14165 // CHECK26-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
14166 // CHECK26-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
14167 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
14168 // CHECK26-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
14169 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
14170 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
14171 // CHECK26-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
14172 // CHECK26-NEXT:    [[CONV3:%.*]] = sitofp i32 [[TMP4]] to double
14173 // CHECK26-NEXT:    [[ADD:%.*]] = fadd double [[CONV3]], 1.500000e+00
14174 // CHECK26-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
14175 // CHECK26-NEXT:    store double [[ADD]], double* [[A]], align 8
14176 // CHECK26-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
14177 // CHECK26-NEXT:    [[TMP5:%.*]] = load double, double* [[A4]], align 8
14178 // CHECK26-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
14179 // CHECK26-NEXT:    store double [[INC]], double* [[A4]], align 8
14180 // CHECK26-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
14181 // CHECK26-NEXT:    [[TMP6:%.*]] = mul nsw i64 1, [[TMP2]]
14182 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP6]]
14183 // CHECK26-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
14184 // CHECK26-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2
14185 // CHECK26-NEXT:    ret void
14186 //
14187 //
14188 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
14189 // CHECK26-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14190 // CHECK26-NEXT:  entry:
14191 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
14192 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
14193 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
14194 // CHECK26-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
14195 // CHECK26-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
14196 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
14197 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
14198 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
14199 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
14200 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
14201 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
14202 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
14203 // CHECK26-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
14204 // CHECK26-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
14205 // CHECK26-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
14206 // CHECK26-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
14207 // CHECK26-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
14208 // CHECK26-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
14209 // CHECK26-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
14210 // CHECK26-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
14211 // CHECK26-NEXT:    ret void
14212 //
14213 //
14214 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..11
14215 // 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]] {
14216 // CHECK26-NEXT:  entry:
14217 // CHECK26-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
14218 // CHECK26-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
14219 // CHECK26-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
14220 // CHECK26-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
14221 // CHECK26-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
14222 // CHECK26-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
14223 // CHECK26-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
14224 // CHECK26-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
14225 // CHECK26-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
14226 // CHECK26-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
14227 // CHECK26-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
14228 // CHECK26-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
14229 // CHECK26-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
14230 // CHECK26-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
14231 // CHECK26-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
14232 // CHECK26-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4
14233 // CHECK26-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
14234 // CHECK26-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
14235 // CHECK26-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
14236 // CHECK26-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
14237 // CHECK26-NEXT:    store i16 [[CONV4]], i16* [[CONV1]], align 2
14238 // CHECK26-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
14239 // CHECK26-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
14240 // CHECK26-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP3]], 1
14241 // CHECK26-NEXT:    store i32 [[ADD5]], i32* [[ARRAYIDX]], align 4
14242 // CHECK26-NEXT:    ret void
14243 //
14244 //
14245 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
14246 // CHECK27-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
14247 // CHECK27-NEXT:  entry:
14248 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14249 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14250 // CHECK27-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
14251 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14252 // CHECK27-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
14253 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14254 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14255 // CHECK27-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14256 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14257 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14258 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14259 // CHECK27-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
14260 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
14261 // CHECK27-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14262 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
14263 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14264 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
14265 // CHECK27-NEXT:    ret void
14266 //
14267 //
14268 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
14269 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14270 // CHECK27-NEXT:  entry:
14271 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14272 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14273 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14274 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14275 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14276 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14277 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14278 // CHECK27-NEXT:    ret void
14279 //
14280 //
14281 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
14282 // CHECK27-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14283 // CHECK27-NEXT:  entry:
14284 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14285 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14286 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14287 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14288 // CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
14289 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14290 // CHECK27-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
14291 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14292 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14293 // CHECK27-NEXT:    ret void
14294 //
14295 //
14296 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
14297 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14298 // CHECK27-NEXT:  entry:
14299 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14300 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14301 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14302 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14303 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14304 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14305 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14306 // CHECK27-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
14307 // CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
14308 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
14309 // CHECK27-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
14310 // CHECK27-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
14311 // CHECK27-NEXT:    ret void
14312 //
14313 //
14314 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
14315 // CHECK27-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14316 // CHECK27-NEXT:  entry:
14317 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14318 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14319 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14320 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14321 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14322 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14323 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14324 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14325 // CHECK27-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
14326 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
14327 // CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
14328 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14329 // CHECK27-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
14330 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14331 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
14332 // CHECK27-NEXT:    ret void
14333 //
14334 //
14335 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..2
14336 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14337 // CHECK27-NEXT:  entry:
14338 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14339 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14340 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14341 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14342 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14343 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14344 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14345 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14346 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14347 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14348 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
14349 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14350 // CHECK27-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
14351 // CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
14352 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
14353 // CHECK27-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
14354 // CHECK27-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
14355 // CHECK27-NEXT:    ret void
14356 //
14357 //
14358 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
14359 // 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:%.*]]) #[[ATTR0]] {
14360 // CHECK27-NEXT:  entry:
14361 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14362 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14363 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14364 // CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14365 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14366 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14367 // CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14368 // CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14369 // CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14370 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14371 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14372 // CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14373 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14374 // CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14375 // CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14376 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14377 // CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14378 // CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14379 // CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14380 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14381 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14382 // CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14383 // CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14384 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14385 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14386 // CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14387 // CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14388 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14389 // CHECK27-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
14390 // CHECK27-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
14391 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
14392 // CHECK27-NEXT:    ret void
14393 //
14394 //
14395 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..3
14396 // 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:%.*]]) #[[ATTR0]] {
14397 // CHECK27-NEXT:  entry:
14398 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14399 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14400 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14401 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14402 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14403 // CHECK27-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14404 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14405 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14406 // CHECK27-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14407 // CHECK27-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14408 // CHECK27-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14409 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14410 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14411 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14412 // CHECK27-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14413 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14414 // CHECK27-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14415 // CHECK27-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14416 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14417 // CHECK27-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14418 // CHECK27-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14419 // CHECK27-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14420 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14421 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14422 // CHECK27-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14423 // CHECK27-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14424 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14425 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14426 // CHECK27-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14427 // CHECK27-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14428 // CHECK27-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14429 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
14430 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14431 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
14432 // CHECK27-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
14433 // CHECK27-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
14434 // CHECK27-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
14435 // CHECK27-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
14436 // CHECK27-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
14437 // CHECK27-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
14438 // CHECK27-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
14439 // CHECK27-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
14440 // CHECK27-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
14441 // CHECK27-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
14442 // CHECK27-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
14443 // CHECK27-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
14444 // CHECK27-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
14445 // CHECK27-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
14446 // CHECK27-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
14447 // CHECK27-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
14448 // CHECK27-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
14449 // CHECK27-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
14450 // CHECK27-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
14451 // CHECK27-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
14452 // CHECK27-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
14453 // CHECK27-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
14454 // CHECK27-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
14455 // CHECK27-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
14456 // CHECK27-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
14457 // CHECK27-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
14458 // CHECK27-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
14459 // CHECK27-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
14460 // CHECK27-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
14461 // CHECK27-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
14462 // CHECK27-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
14463 // CHECK27-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
14464 // CHECK27-NEXT:    ret void
14465 //
14466 //
14467 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
14468 // CHECK27-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14469 // CHECK27-NEXT:  entry:
14470 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14471 // CHECK27-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
14472 // CHECK27-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14473 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
14474 // CHECK27-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
14475 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
14476 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14477 // CHECK27-NEXT:    ret void
14478 //
14479 //
14480 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..4
14481 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14482 // CHECK27-NEXT:  entry:
14483 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14484 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14485 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14486 // CHECK27-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
14487 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14488 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14489 // CHECK27-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14490 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
14491 // CHECK27-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
14492 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
14493 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14494 // CHECK27-NEXT:    ret void
14495 //
14496 //
14497 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..5
14498 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14499 // CHECK27-NEXT:  entry:
14500 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14501 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14502 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14503 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14504 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14505 // CHECK27-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14506 // CHECK27-NEXT:    ret void
14507 //
14508 //
14509 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
14510 // CHECK27-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14511 // CHECK27-NEXT:  entry:
14512 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14513 // CHECK27-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
14514 // CHECK27-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14515 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
14516 // CHECK27-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
14517 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
14518 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14519 // CHECK27-NEXT:    ret void
14520 //
14521 //
14522 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..6
14523 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14524 // CHECK27-NEXT:  entry:
14525 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14526 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14527 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14528 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14529 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14530 // CHECK27-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14531 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
14532 // CHECK27-NEXT:    ret void
14533 //
14534 //
14535 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..7
14536 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
14537 // CHECK27-NEXT:  entry:
14538 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14539 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14540 // CHECK27-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
14541 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14542 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14543 // CHECK27-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
14544 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
14545 // CHECK27-NEXT:    ret void
14546 //
14547 //
14548 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
14549 // CHECK27-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
14550 // CHECK27-NEXT:  entry:
14551 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14552 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14553 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14554 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32 [[TMP0]])
14555 // CHECK27-NEXT:    ret void
14556 //
14557 //
14558 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..8
14559 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
14560 // CHECK27-NEXT:  entry:
14561 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14562 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14563 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14564 // CHECK27-NEXT:    [[F:%.*]] = alloca i32*, align 4
14565 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14566 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14567 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14568 // CHECK27-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14569 // CHECK27-NEXT:    ret void
14570 //
14571 //
14572 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
14573 // CHECK27-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14574 // CHECK27-NEXT:  entry:
14575 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14576 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14577 // CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
14578 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
14579 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14580 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14581 // CHECK27-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
14582 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14583 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14584 // CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
14585 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
14586 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14587 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
14588 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
14589 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
14590 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
14591 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
14592 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
14593 // CHECK27-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14594 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
14595 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14596 // CHECK27-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
14597 // CHECK27-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
14598 // CHECK27-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
14599 // CHECK27-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
14600 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
14601 // CHECK27-NEXT:    ret void
14602 //
14603 //
14604 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..9
14605 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14606 // CHECK27-NEXT:  entry:
14607 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14608 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14609 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14610 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14611 // CHECK27-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
14612 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
14613 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14614 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14615 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14616 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14617 // CHECK27-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
14618 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
14619 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14620 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
14621 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
14622 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
14623 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
14624 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14625 // CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
14626 // CHECK27-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
14627 // CHECK27-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
14628 // CHECK27-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
14629 // CHECK27-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
14630 // CHECK27-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
14631 // CHECK27-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
14632 // CHECK27-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
14633 // CHECK27-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
14634 // CHECK27-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
14635 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
14636 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
14637 // CHECK27-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
14638 // CHECK27-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
14639 // CHECK27-NEXT:    ret void
14640 //
14641 //
14642 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
14643 // CHECK27-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
14644 // CHECK27-NEXT:  entry:
14645 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
14646 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
14647 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14648 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14649 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
14650 // CHECK27-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
14651 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
14652 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
14653 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14654 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14655 // CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
14656 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
14657 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14658 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14659 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
14660 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
14661 // CHECK27-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
14662 // CHECK27-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
14663 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
14664 // CHECK27-NEXT:    ret void
14665 //
14666 //
14667 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..10
14668 // 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]] {
14669 // CHECK27-NEXT:  entry:
14670 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14671 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14672 // CHECK27-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
14673 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
14674 // CHECK27-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14675 // CHECK27-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14676 // CHECK27-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
14677 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14678 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14679 // CHECK27-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
14680 // CHECK27-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
14681 // CHECK27-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14682 // CHECK27-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14683 // CHECK27-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
14684 // CHECK27-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
14685 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14686 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14687 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
14688 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
14689 // CHECK27-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
14690 // CHECK27-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
14691 // CHECK27-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
14692 // CHECK27-NEXT:    store double [[ADD]], double* [[A]], align 4
14693 // CHECK27-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
14694 // CHECK27-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
14695 // CHECK27-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
14696 // CHECK27-NEXT:    store double [[INC]], double* [[A3]], align 4
14697 // CHECK27-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
14698 // CHECK27-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
14699 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
14700 // CHECK27-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
14701 // CHECK27-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
14702 // CHECK27-NEXT:    ret void
14703 //
14704 //
14705 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
14706 // CHECK27-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
14707 // CHECK27-NEXT:  entry:
14708 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14709 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14710 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
14711 // CHECK27-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14712 // CHECK27-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14713 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14714 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14715 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
14716 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14717 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
14718 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
14719 // CHECK27-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
14720 // CHECK27-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
14721 // CHECK27-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
14722 // CHECK27-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14723 // CHECK27-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
14724 // CHECK27-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14725 // CHECK27-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
14726 // CHECK27-NEXT:    ret void
14727 //
14728 //
14729 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..11
14730 // 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]] {
14731 // CHECK27-NEXT:  entry:
14732 // CHECK27-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14733 // CHECK27-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14734 // CHECK27-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14735 // CHECK27-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14736 // CHECK27-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
14737 // CHECK27-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14738 // CHECK27-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14739 // CHECK27-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14740 // CHECK27-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14741 // CHECK27-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
14742 // CHECK27-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14743 // CHECK27-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
14744 // CHECK27-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
14745 // CHECK27-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
14746 // CHECK27-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14747 // CHECK27-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
14748 // CHECK27-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
14749 // CHECK27-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
14750 // CHECK27-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
14751 // CHECK27-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
14752 // CHECK27-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
14753 // CHECK27-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
14754 // CHECK27-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
14755 // CHECK27-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
14756 // CHECK27-NEXT:    ret void
14757 //
14758 //
14759 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
14760 // CHECK28-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]], i32 noundef [[DOTCAPTURE_EXPR_1:%.*]]) #[[ATTR0:[0-9]+]] {
14761 // CHECK28-NEXT:  entry:
14762 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14763 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14764 // CHECK28-NEXT:    [[DOTCAPTURE_EXPR__ADDR2:%.*]] = alloca i32, align 4
14765 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14766 // CHECK28-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB1:[0-9]+]])
14767 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14768 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14769 // CHECK28-NEXT:    store i32 [[DOTCAPTURE_EXPR_1]], i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14770 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14771 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14772 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR2]], align 4
14773 // CHECK28-NEXT:    call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]])
14774 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
14775 // CHECK28-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14776 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
14777 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14778 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]])
14779 // CHECK28-NEXT:    ret void
14780 //
14781 //
14782 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
14783 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14784 // CHECK28-NEXT:  entry:
14785 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14786 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14787 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14788 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14789 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14790 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14791 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14792 // CHECK28-NEXT:    ret void
14793 //
14794 //
14795 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l111
14796 // CHECK28-SAME: (i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14797 // CHECK28-NEXT:  entry:
14798 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14799 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14800 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14801 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14802 // CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
14803 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14804 // CHECK28-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
14805 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14806 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14807 // CHECK28-NEXT:    ret void
14808 //
14809 //
14810 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
14811 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14812 // CHECK28-NEXT:  entry:
14813 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14814 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14815 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14816 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14817 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14818 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14819 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14820 // CHECK28-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
14821 // CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP0]] to i32
14822 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV1]], 1
14823 // CHECK28-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
14824 // CHECK28-NEXT:    store i16 [[CONV2]], i16* [[CONV]], align 2
14825 // CHECK28-NEXT:    ret void
14826 //
14827 //
14828 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l118
14829 // CHECK28-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14830 // CHECK28-NEXT:  entry:
14831 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14832 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14833 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14834 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
14835 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14836 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14837 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14838 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14839 // CHECK28-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
14840 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
14841 // CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
14842 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
14843 // CHECK28-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
14844 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
14845 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
14846 // CHECK28-NEXT:    ret void
14847 //
14848 //
14849 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..2
14850 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
14851 // CHECK28-NEXT:  entry:
14852 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14853 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14854 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14855 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
14856 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14857 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14858 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14859 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
14860 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
14861 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
14862 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
14863 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14864 // CHECK28-NEXT:    [[TMP1:%.*]] = load i16, i16* [[CONV]], align 2
14865 // CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP1]] to i32
14866 // CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
14867 // CHECK28-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
14868 // CHECK28-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
14869 // CHECK28-NEXT:    ret void
14870 //
14871 //
14872 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l142
14873 // 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:%.*]]) #[[ATTR0]] {
14874 // CHECK28-NEXT:  entry:
14875 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14876 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14877 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14878 // CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14879 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14880 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14881 // CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14882 // CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14883 // CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14884 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
14885 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14886 // CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14887 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14888 // CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14889 // CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14890 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14891 // CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14892 // CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14893 // CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14894 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14895 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14896 // CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14897 // CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14898 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14899 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14900 // CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14901 // CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14902 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14903 // CHECK28-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
14904 // CHECK28-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
14905 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 9, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*)* @.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]])
14906 // CHECK28-NEXT:    ret void
14907 //
14908 //
14909 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..3
14910 // 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:%.*]]) #[[ATTR0]] {
14911 // CHECK28-NEXT:  entry:
14912 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14913 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14914 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
14915 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
14916 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
14917 // CHECK28-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
14918 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
14919 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
14920 // CHECK28-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
14921 // CHECK28-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
14922 // CHECK28-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
14923 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14924 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14925 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
14926 // CHECK28-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
14927 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
14928 // CHECK28-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
14929 // CHECK28-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
14930 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
14931 // CHECK28-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
14932 // CHECK28-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
14933 // CHECK28-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
14934 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
14935 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
14936 // CHECK28-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
14937 // CHECK28-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
14938 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
14939 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
14940 // CHECK28-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
14941 // CHECK28-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
14942 // CHECK28-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
14943 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP8]], 1
14944 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
14945 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
14946 // CHECK28-NEXT:    [[TMP9:%.*]] = load float, float* [[ARRAYIDX]], align 4
14947 // CHECK28-NEXT:    [[CONV:%.*]] = fpext float [[TMP9]] to double
14948 // CHECK28-NEXT:    [[ADD5:%.*]] = fadd double [[CONV]], 1.000000e+00
14949 // CHECK28-NEXT:    [[CONV6:%.*]] = fptrunc double [[ADD5]] to float
14950 // CHECK28-NEXT:    store float [[CONV6]], float* [[ARRAYIDX]], align 4
14951 // CHECK28-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
14952 // CHECK28-NEXT:    [[TMP10:%.*]] = load float, float* [[ARRAYIDX7]], align 4
14953 // CHECK28-NEXT:    [[CONV8:%.*]] = fpext float [[TMP10]] to double
14954 // CHECK28-NEXT:    [[ADD9:%.*]] = fadd double [[CONV8]], 1.000000e+00
14955 // CHECK28-NEXT:    [[CONV10:%.*]] = fptrunc double [[ADD9]] to float
14956 // CHECK28-NEXT:    store float [[CONV10]], float* [[ARRAYIDX7]], align 4
14957 // CHECK28-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
14958 // CHECK28-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX11]], i32 0, i32 2
14959 // CHECK28-NEXT:    [[TMP11:%.*]] = load double, double* [[ARRAYIDX12]], align 8
14960 // CHECK28-NEXT:    [[ADD13:%.*]] = fadd double [[TMP11]], 1.000000e+00
14961 // CHECK28-NEXT:    store double [[ADD13]], double* [[ARRAYIDX12]], align 8
14962 // CHECK28-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP5]]
14963 // CHECK28-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP12]]
14964 // CHECK28-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX14]], i32 3
14965 // CHECK28-NEXT:    [[TMP13:%.*]] = load double, double* [[ARRAYIDX15]], align 8
14966 // CHECK28-NEXT:    [[ADD16:%.*]] = fadd double [[TMP13]], 1.000000e+00
14967 // CHECK28-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8
14968 // CHECK28-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
14969 // CHECK28-NEXT:    [[TMP14:%.*]] = load i64, i64* [[X]], align 4
14970 // CHECK28-NEXT:    [[ADD17:%.*]] = add nsw i64 [[TMP14]], 1
14971 // CHECK28-NEXT:    store i64 [[ADD17]], i64* [[X]], align 4
14972 // CHECK28-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
14973 // CHECK28-NEXT:    [[TMP15:%.*]] = load i8, i8* [[Y]], align 4
14974 // CHECK28-NEXT:    [[CONV18:%.*]] = sext i8 [[TMP15]] to i32
14975 // CHECK28-NEXT:    [[ADD19:%.*]] = add nsw i32 [[CONV18]], 1
14976 // CHECK28-NEXT:    [[CONV20:%.*]] = trunc i32 [[ADD19]] to i8
14977 // CHECK28-NEXT:    store i8 [[CONV20]], i8* [[Y]], align 4
14978 // CHECK28-NEXT:    ret void
14979 //
14980 //
14981 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l154
14982 // CHECK28-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14983 // CHECK28-NEXT:  entry:
14984 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
14985 // CHECK28-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
14986 // CHECK28-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
14987 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
14988 // CHECK28-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
14989 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
14990 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]])
14991 // CHECK28-NEXT:    ret void
14992 //
14993 //
14994 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..4
14995 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
14996 // CHECK28-NEXT:  entry:
14997 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14998 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14999 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
15000 // CHECK28-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
15001 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15002 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15003 // CHECK28-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
15004 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
15005 // CHECK28-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
15006 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
15007 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32 [[TMP1]])
15008 // CHECK28-NEXT:    ret void
15009 //
15010 //
15011 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..5
15012 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
15013 // CHECK28-NEXT:  entry:
15014 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15015 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15016 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
15017 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15018 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15019 // CHECK28-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
15020 // CHECK28-NEXT:    ret void
15021 //
15022 //
15023 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l157
15024 // CHECK28-SAME: (i32 noundef [[NN:%.*]]) #[[ATTR0]] {
15025 // CHECK28-NEXT:  entry:
15026 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
15027 // CHECK28-NEXT:    [[NN_CASTED:%.*]] = alloca i32, align 4
15028 // CHECK28-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
15029 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[NN_ADDR]], align 4
15030 // CHECK28-NEXT:    store i32 [[TMP0]], i32* [[NN_CASTED]], align 4
15031 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[NN_CASTED]], align 4
15032 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP1]])
15033 // CHECK28-NEXT:    ret void
15034 //
15035 //
15036 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..6
15037 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[NN:%.*]]) #[[ATTR0]] {
15038 // CHECK28-NEXT:  entry:
15039 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15040 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15041 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32, align 4
15042 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15043 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15044 // CHECK28-NEXT:    store i32 [[NN]], i32* [[NN_ADDR]], align 4
15045 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32* [[NN_ADDR]])
15046 // CHECK28-NEXT:    ret void
15047 //
15048 //
15049 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..7
15050 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[NN:%.*]]) #[[ATTR0]] {
15051 // CHECK28-NEXT:  entry:
15052 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15053 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15054 // CHECK28-NEXT:    [[NN_ADDR:%.*]] = alloca i32*, align 4
15055 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15056 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15057 // CHECK28-NEXT:    store i32* [[NN]], i32** [[NN_ADDR]], align 4
15058 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[NN_ADDR]], align 4
15059 // CHECK28-NEXT:    ret void
15060 //
15061 //
15062 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z6bazzzziPi_l182
15063 // CHECK28-SAME: (i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
15064 // CHECK28-NEXT:  entry:
15065 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15066 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15067 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15068 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32 [[TMP0]])
15069 // CHECK28-NEXT:    ret void
15070 //
15071 //
15072 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..8
15073 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[VLA:%.*]]) #[[ATTR0]] {
15074 // CHECK28-NEXT:  entry:
15075 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15076 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15077 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15078 // CHECK28-NEXT:    [[F:%.*]] = alloca i32*, align 4
15079 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15080 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15081 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15082 // CHECK28-NEXT:    [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15083 // CHECK28-NEXT:    ret void
15084 //
15085 //
15086 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l209
15087 // CHECK28-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
15088 // CHECK28-NEXT:  entry:
15089 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15090 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15091 // CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
15092 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15093 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15094 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15095 // CHECK28-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
15096 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15097 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15098 // CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
15099 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15100 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15101 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
15102 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15103 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15104 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
15105 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
15106 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
15107 // CHECK28-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15108 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
15109 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15110 // CHECK28-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
15111 // CHECK28-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
15112 // CHECK28-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
15113 // CHECK28-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
15114 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
15115 // CHECK28-NEXT:    ret void
15116 //
15117 //
15118 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..9
15119 // CHECK28-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
15120 // CHECK28-NEXT:  entry:
15121 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15122 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15123 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15124 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15125 // CHECK28-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
15126 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15127 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15128 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15129 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15130 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15131 // CHECK28-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
15132 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15133 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15134 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
15135 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15136 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15137 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
15138 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
15139 // CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
15140 // CHECK28-NEXT:    [[CONV2:%.*]] = sext i16 [[TMP2]] to i32
15141 // CHECK28-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV2]], 1
15142 // CHECK28-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i16
15143 // CHECK28-NEXT:    store i16 [[CONV4]], i16* [[CONV]], align 2
15144 // CHECK28-NEXT:    [[TMP3:%.*]] = load i8, i8* [[CONV1]], align 1
15145 // CHECK28-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP3]] to i32
15146 // CHECK28-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
15147 // CHECK28-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i8
15148 // CHECK28-NEXT:    store i8 [[CONV7]], i8* [[CONV1]], align 1
15149 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
15150 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
15151 // CHECK28-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP4]], 1
15152 // CHECK28-NEXT:    store i32 [[ADD8]], i32* [[ARRAYIDX]], align 4
15153 // CHECK28-NEXT:    ret void
15154 //
15155 //
15156 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l227
15157 // CHECK28-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
15158 // CHECK28-NEXT:  entry:
15159 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
15160 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
15161 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15162 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
15163 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
15164 // CHECK28-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
15165 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
15166 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
15167 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15168 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
15169 // CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
15170 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
15171 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15172 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
15173 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
15174 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
15175 // CHECK28-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
15176 // CHECK28-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
15177 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
15178 // CHECK28-NEXT:    ret void
15179 //
15180 //
15181 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..10
15182 // 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]] {
15183 // CHECK28-NEXT:  entry:
15184 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15185 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15186 // CHECK28-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
15187 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
15188 // CHECK28-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
15189 // CHECK28-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
15190 // CHECK28-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
15191 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15192 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15193 // CHECK28-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
15194 // CHECK28-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
15195 // CHECK28-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15196 // CHECK28-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
15197 // CHECK28-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
15198 // CHECK28-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
15199 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15200 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
15201 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
15202 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
15203 // CHECK28-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP4]] to double
15204 // CHECK28-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
15205 // CHECK28-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
15206 // CHECK28-NEXT:    store double [[ADD]], double* [[A]], align 4
15207 // CHECK28-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
15208 // CHECK28-NEXT:    [[TMP5:%.*]] = load double, double* [[A3]], align 4
15209 // CHECK28-NEXT:    [[INC:%.*]] = fadd double [[TMP5]], 1.000000e+00
15210 // CHECK28-NEXT:    store double [[INC]], double* [[A3]], align 4
15211 // CHECK28-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
15212 // CHECK28-NEXT:    [[TMP6:%.*]] = mul nsw i32 1, [[TMP2]]
15213 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP6]]
15214 // CHECK28-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
15215 // CHECK28-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2
15216 // CHECK28-NEXT:    ret void
15217 //
15218 //
15219 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l192
15220 // CHECK28-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
15221 // CHECK28-NEXT:  entry:
15222 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15223 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15224 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15225 // CHECK28-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
15226 // CHECK28-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
15227 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15228 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15229 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15230 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15231 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15232 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15233 // CHECK28-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
15234 // CHECK28-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
15235 // CHECK28-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
15236 // CHECK28-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
15237 // CHECK28-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
15238 // CHECK28-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
15239 // CHECK28-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
15240 // CHECK28-NEXT:    ret void
15241 //
15242 //
15243 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..11
15244 // 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]] {
15245 // CHECK28-NEXT:  entry:
15246 // CHECK28-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15247 // CHECK28-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15248 // CHECK28-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
15249 // CHECK28-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
15250 // CHECK28-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
15251 // CHECK28-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15252 // CHECK28-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15253 // CHECK28-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
15254 // CHECK28-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
15255 // CHECK28-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
15256 // CHECK28-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
15257 // CHECK28-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
15258 // CHECK28-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
15259 // CHECK28-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 1
15260 // CHECK28-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4
15261 // CHECK28-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
15262 // CHECK28-NEXT:    [[CONV1:%.*]] = sext i16 [[TMP2]] to i32
15263 // CHECK28-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV1]], 1
15264 // CHECK28-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
15265 // CHECK28-NEXT:    store i16 [[CONV3]], i16* [[CONV]], align 2
15266 // CHECK28-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
15267 // CHECK28-NEXT:    [[TMP3:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
15268 // CHECK28-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP3]], 1
15269 // CHECK28-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4
15270 // CHECK28-NEXT:    ret void
15271 //
15272