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=CHECK1
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=CHECK3
9 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK5
10 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
11 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK5
12 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK7
13 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
14 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK7
15 
16 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
17 // 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
18 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
19 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
20 // 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
21 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
22 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK13
23 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
24 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK13
25 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck -allow-deprecated-dag-overlap  %s --check-prefix=CHECK15
26 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
27 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK15
28 
29 // Test target codegen - host bc file has to be created first.
30 // 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
31 // 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=CHECK17
32 // 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
33 // 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=CHECK17
34 // 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
35 // 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=CHECK19
36 // 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
37 // 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=CHECK19
38 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
39 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK21
40 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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
41 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK21
42 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
43 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK23
44 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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
45 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=50 -DOMP5 -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=CHECK23
46 
47 // 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
48 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
49 // 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
50 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK9
51 // 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
52 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
53 // 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
54 // 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 -allow-deprecated-dag-overlap  %s --check-prefix=CHECK11
55 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm-bc %s -o %t-ppc-host.bc
56 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK13
57 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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
58 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK13
59 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm-bc %s -o %t-x86-host.bc
60 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK15
61 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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
62 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=50 -DOMP5 -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=CHECK15
63 
64 // expected-no-diagnostics
65 #ifndef HEADER
66 #define HEADER
67 
68 
69 
70 // We have 8 target regions, but only 7 that actually will generate offloading
71 // code, only 6 will have mapped arguments, and only 4 have all-constant map
72 // sizes.
73 
74 
75 
76 // Check target registration is registered as a Ctor.
77 
78 
79 template<typename tx, typename ty>
80 struct TT{
81   tx X;
82   ty Y;
83 };
84 
85 long long get_val() { return 0; }
86 
87 int foo(int n) {
88   int a = 0;
89   short aa = 0;
90   float b[10];
91   float bn[n];
92   double c[5][10];
93   double cn[5][n];
94   TT<long long, char> d;
95 
96   #pragma omp target parallel for simd nowait
97   for (int i = 3; i < 32; i += 5) {
98   }
99 
100   long long k = get_val();
101   #pragma omp target parallel for simd if(target: 0) linear(k : 3) schedule(dynamic)
102   for (int i = 10; i > 1; i--) {
103     a += 1;
104   }
105 
106 
107   int lin = 12;
108   #pragma omp target parallel for simd if(target: 1) linear(lin, a : get_val())
109   for (unsigned long long it = 2000; it >= 600; it-=400) {
110     aa += 1;
111   }
112 
113 
114 
115 
116   #pragma omp target parallel for simd if(target: n>10)
117   for (short it = 6; it <= 20; it-=-4) {
118     a += 1;
119     aa += 1;
120   }
121 
122   // We capture 3 VLA sizes in this target region
123 
124 
125 
126 
127 
128   // The names below are not necessarily consistent with the names used for the
129   // addresses above as some are repeated.
130 
131 
132 
133 
134 
135 
136 
137 
138 
139 
140   #pragma omp target parallel for simd if(target: n>20) schedule(static, a)
141   for (unsigned char it = 'z'; it >= 'a'; it+=-1) {
142     a += 1;
143     b[2] += 1.0;
144     bn[3] += 1.0;
145     c[1][2] += 1.0;
146     cn[1][3] += 1.0;
147     d.X += 1;
148     d.Y += 1;
149   }
150 
151   return a;
152 }
153 
154 // Check that the offloading functions are emitted and that the arguments are
155 // correct and loaded correctly for the target regions in foo().
156 
157 
158 
159 
160 // Create stack storage and store argument in there.
161 
162 // Create stack storage and store argument in there.
163 
164 // Create stack storage and store argument in there.
165 
166 // Create local storage for each capture.
167 
168 
169 
170 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
171 
172 template<typename tx>
173 tx ftemplate(int n) {
174   tx a = 0;
175   short aa = 0;
176   tx b[10];
177 
178   #pragma omp target parallel for simd if(target: n>40)
179   for (long long i = -10; i < 10; i += 3) {
180     a += 1;
181     aa += 1;
182     b[2] += 1;
183   }
184 
185   return a;
186 }
187 
188 static
189 int fstatic(int n) {
190   int a = 0;
191   short aa = 0;
192   char aaa = 0;
193   int b[10];
194 
195   #pragma omp target parallel for simd if(target: n>50)
196   for (unsigned i=100; i<10; i+=10) {
197     a += 1;
198     aa += 1;
199     aaa += 1;
200     b[2] += 1;
201   }
202 
203   return a;
204 }
205 
206 struct S1 {
207   double a;
208 
209   int r1(int n){
210     int b = n+1;
211     short int c[2][n];
212 
213 #ifdef OMP5
214     #pragma omp target parallel for simd if(n>60) nontemporal(a)
215 #else
216     #pragma omp target parallel for simd if(target: n>60)
217 #endif // OMP5
218     for (unsigned long long it = 2000; it >= 600; it -= 400) {
219       this->a = (double)b + 1.5;
220       c[1][1] = ++a;
221     }
222 
223     return c[1][1] + (int)b;
224   }
225 };
226 
227 int bar(int n){
228   int a = 0;
229 
230   a += foo(n);
231 
232   S1 S;
233   a += S.r1(n);
234 
235   a += fstatic(n);
236 
237   a += ftemplate<int>(n);
238 
239   return a;
240 }
241 
242 
243 
244 // We capture 2 VLA sizes in this target region
245 
246 
247 // The names below are not necessarily consistent with the names used for the
248 // addresses above as some are repeated.
249 
250 
251 
252 
253 
254 
255 
256 
257 
258 
259 
260 
261 
262 
263 
264 
265 
266 
267 
268 // Check that the offloading functions are emitted and that the arguments are
269 // correct and loaded correctly for the target regions of the callees of bar().
270 
271 // Create local storage for each capture.
272 // Store captures in the context.
273 
274 
275 
276 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
277 
278 
279 // Create local storage for each capture.
280 // Store captures in the context.
281 
282 
283 
284 
285 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
286 
287 // Create local storage for each capture.
288 // Store captures in the context.
289 
290 
291 
292 // To reduce complexity, we're only going as far as validating the signature of the outlined parallel function.
293 
294 
295 #endif
296 // CHECK1-LABEL: define {{[^@]+}}@_Z7get_valv
297 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
298 // CHECK1-NEXT:  entry:
299 // CHECK1-NEXT:    ret i64 0
300 //
301 //
302 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooi
303 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
304 // CHECK1-NEXT:  entry:
305 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
306 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
307 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
308 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
309 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
310 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
311 // CHECK1-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
312 // CHECK1-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
313 // CHECK1-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
314 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
315 // CHECK1-NEXT:    [[K:%.*]] = alloca i64, align 8
316 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
317 // CHECK1-NEXT:    [[K_CASTED:%.*]] = alloca i64, align 8
318 // CHECK1-NEXT:    [[LIN:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
320 // CHECK1-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
321 // CHECK1-NEXT:    [[A_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:    [[A_CASTED6:%.*]] = alloca i64, align 8
326 // CHECK1-NEXT:    [[AA_CASTED8:%.*]] = alloca i64, align 8
327 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [2 x i8*], align 8
328 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [2 x i8*], align 8
329 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [2 x i8*], align 8
330 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
331 // CHECK1-NEXT:    [[A_CASTED15:%.*]] = alloca i64, align 8
332 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
333 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [10 x i8*], align 8
334 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [10 x i8*], align 8
335 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [10 x i8*], align 8
336 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
337 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
338 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
339 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
340 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
341 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
342 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
343 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
344 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
345 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
346 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
347 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
348 // CHECK1-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
349 // CHECK1-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
350 // CHECK1-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
351 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
352 // CHECK1-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
353 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
354 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
355 // CHECK1-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP7]])
356 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
357 // CHECK1-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
358 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
359 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
360 // CHECK1-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
361 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
362 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[K]], align 8
363 // CHECK1-NEXT:    store i64 [[TMP13]], i64* [[K_CASTED]], align 8
364 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[K_CASTED]], align 8
365 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP12]], i64 [[TMP14]]) #[[ATTR4:[0-9]+]]
366 // CHECK1-NEXT:    store i32 12, i32* [[LIN]], align 4
367 // CHECK1-NEXT:    [[TMP15:%.*]] = load i16, i16* [[AA]], align 2
368 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
369 // CHECK1-NEXT:    store i16 [[TMP15]], i16* [[CONV2]], align 2
370 // CHECK1-NEXT:    [[TMP16:%.*]] = load i64, i64* [[AA_CASTED]], align 8
371 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[LIN]], align 4
372 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
373 // CHECK1-NEXT:    store i32 [[TMP17]], i32* [[CONV3]], align 4
374 // CHECK1-NEXT:    [[TMP18:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
375 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
376 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED4]] to i32*
377 // CHECK1-NEXT:    store i32 [[TMP19]], i32* [[CONV5]], align 4
378 // CHECK1-NEXT:    [[TMP20:%.*]] = load i64, i64* [[A_CASTED4]], align 8
379 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
380 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
381 // CHECK1-NEXT:    store i64 [[TMP16]], i64* [[TMP22]], align 8
382 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
383 // CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
384 // CHECK1-NEXT:    store i64 [[TMP16]], i64* [[TMP24]], align 8
385 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
386 // CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
387 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
388 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
389 // CHECK1-NEXT:    store i64 [[TMP18]], i64* [[TMP27]], align 8
390 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
391 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
392 // CHECK1-NEXT:    store i64 [[TMP18]], i64* [[TMP29]], align 8
393 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
394 // CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
395 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
396 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
397 // CHECK1-NEXT:    store i64 [[TMP20]], i64* [[TMP32]], align 8
398 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
399 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i64*
400 // CHECK1-NEXT:    store i64 [[TMP20]], i64* [[TMP34]], align 8
401 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
402 // CHECK1-NEXT:    store i8* null, i8** [[TMP35]], align 8
403 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
404 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
405 // CHECK1-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108.region_id, i32 3, i8** [[TMP36]], i8** [[TMP37]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
406 // CHECK1-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
407 // CHECK1-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
408 // CHECK1:       omp_offload.failed:
409 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108(i64 [[TMP16]], i64 [[TMP18]], i64 [[TMP20]]) #[[ATTR4]]
410 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
411 // CHECK1:       omp_offload.cont:
412 // CHECK1-NEXT:    [[TMP40:%.*]] = load i32, i32* [[A]], align 4
413 // CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[A_CASTED6]] to i32*
414 // CHECK1-NEXT:    store i32 [[TMP40]], i32* [[CONV7]], align 4
415 // CHECK1-NEXT:    [[TMP41:%.*]] = load i64, i64* [[A_CASTED6]], align 8
416 // CHECK1-NEXT:    [[TMP42:%.*]] = load i16, i16* [[AA]], align 2
417 // CHECK1-NEXT:    [[CONV9:%.*]] = bitcast i64* [[AA_CASTED8]] to i16*
418 // CHECK1-NEXT:    store i16 [[TMP42]], i16* [[CONV9]], align 2
419 // CHECK1-NEXT:    [[TMP43:%.*]] = load i64, i64* [[AA_CASTED8]], align 8
420 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[N_ADDR]], align 4
421 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP44]], 10
422 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
423 // CHECK1:       omp_if.then:
424 // CHECK1-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
425 // CHECK1-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
426 // CHECK1-NEXT:    store i64 [[TMP41]], i64* [[TMP46]], align 8
427 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
428 // CHECK1-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
429 // CHECK1-NEXT:    store i64 [[TMP41]], i64* [[TMP48]], align 8
430 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
431 // CHECK1-NEXT:    store i8* null, i8** [[TMP49]], align 8
432 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 1
433 // CHECK1-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
434 // CHECK1-NEXT:    store i64 [[TMP43]], i64* [[TMP51]], align 8
435 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 1
436 // CHECK1-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
437 // CHECK1-NEXT:    store i64 [[TMP43]], i64* [[TMP53]], align 8
438 // CHECK1-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 1
439 // CHECK1-NEXT:    store i8* null, i8** [[TMP54]], align 8
440 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
441 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
442 // CHECK1-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
443 // CHECK1-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
444 // CHECK1-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
445 // CHECK1:       omp_offload.failed13:
446 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i64 [[TMP41]], i64 [[TMP43]]) #[[ATTR4]]
447 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
448 // CHECK1:       omp_offload.cont14:
449 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
450 // CHECK1:       omp_if.else:
451 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i64 [[TMP41]], i64 [[TMP43]]) #[[ATTR4]]
452 // CHECK1-NEXT:    br label [[OMP_IF_END]]
453 // CHECK1:       omp_if.end:
454 // CHECK1-NEXT:    [[TMP59:%.*]] = load i32, i32* [[A]], align 4
455 // CHECK1-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_]], align 4
456 // CHECK1-NEXT:    [[TMP60:%.*]] = load i32, i32* [[A]], align 4
457 // CHECK1-NEXT:    [[CONV16:%.*]] = bitcast i64* [[A_CASTED15]] to i32*
458 // CHECK1-NEXT:    store i32 [[TMP60]], i32* [[CONV16]], align 4
459 // CHECK1-NEXT:    [[TMP61:%.*]] = load i64, i64* [[A_CASTED15]], align 8
460 // CHECK1-NEXT:    [[TMP62:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
461 // CHECK1-NEXT:    [[CONV17:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
462 // CHECK1-NEXT:    store i32 [[TMP62]], i32* [[CONV17]], align 4
463 // CHECK1-NEXT:    [[TMP63:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
464 // CHECK1-NEXT:    [[TMP64:%.*]] = load i32, i32* [[N_ADDR]], align 4
465 // CHECK1-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP64]], 20
466 // CHECK1-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
467 // CHECK1:       omp_if.then19:
468 // CHECK1-NEXT:    [[TMP65:%.*]] = mul nuw i64 [[TMP2]], 4
469 // CHECK1-NEXT:    [[TMP66:%.*]] = mul nuw i64 5, [[TMP5]]
470 // CHECK1-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP66]], 8
471 // CHECK1-NEXT:    [[TMP68:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
472 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP68]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.8 to i8*), i64 80, i1 false)
473 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
474 // CHECK1-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to i64*
475 // CHECK1-NEXT:    store i64 [[TMP61]], i64* [[TMP70]], align 8
476 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
477 // CHECK1-NEXT:    [[TMP72:%.*]] = bitcast i8** [[TMP71]] to i64*
478 // CHECK1-NEXT:    store i64 [[TMP61]], i64* [[TMP72]], align 8
479 // CHECK1-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 0
480 // CHECK1-NEXT:    store i8* null, i8** [[TMP73]], align 8
481 // CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
482 // CHECK1-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [10 x float]**
483 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP75]], align 8
484 // CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
485 // CHECK1-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to [10 x float]**
486 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP77]], align 8
487 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 1
488 // CHECK1-NEXT:    store i8* null, i8** [[TMP78]], align 8
489 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
490 // CHECK1-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
491 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP80]], align 8
492 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
493 // CHECK1-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
494 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP82]], align 8
495 // CHECK1-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 2
496 // CHECK1-NEXT:    store i8* null, i8** [[TMP83]], align 8
497 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
498 // CHECK1-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to float**
499 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP85]], align 8
500 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
501 // CHECK1-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to float**
502 // CHECK1-NEXT:    store float* [[VLA]], float** [[TMP87]], align 8
503 // CHECK1-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
504 // CHECK1-NEXT:    store i64 [[TMP65]], i64* [[TMP88]], align 8
505 // CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 3
506 // CHECK1-NEXT:    store i8* null, i8** [[TMP89]], align 8
507 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
508 // CHECK1-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to [5 x [10 x double]]**
509 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP91]], align 8
510 // CHECK1-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
511 // CHECK1-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to [5 x [10 x double]]**
512 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP93]], align 8
513 // CHECK1-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 4
514 // CHECK1-NEXT:    store i8* null, i8** [[TMP94]], align 8
515 // CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
516 // CHECK1-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
517 // CHECK1-NEXT:    store i64 5, i64* [[TMP96]], align 8
518 // CHECK1-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
519 // CHECK1-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
520 // CHECK1-NEXT:    store i64 5, i64* [[TMP98]], align 8
521 // CHECK1-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 5
522 // CHECK1-NEXT:    store i8* null, i8** [[TMP99]], align 8
523 // CHECK1-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
524 // CHECK1-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i64*
525 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP101]], align 8
526 // CHECK1-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
527 // CHECK1-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i64*
528 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP103]], align 8
529 // CHECK1-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 6
530 // CHECK1-NEXT:    store i8* null, i8** [[TMP104]], align 8
531 // CHECK1-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
532 // CHECK1-NEXT:    [[TMP106:%.*]] = bitcast i8** [[TMP105]] to double**
533 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP106]], align 8
534 // CHECK1-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
535 // CHECK1-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to double**
536 // CHECK1-NEXT:    store double* [[VLA1]], double** [[TMP108]], align 8
537 // CHECK1-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
538 // CHECK1-NEXT:    store i64 [[TMP67]], i64* [[TMP109]], align 8
539 // CHECK1-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 7
540 // CHECK1-NEXT:    store i8* null, i8** [[TMP110]], align 8
541 // CHECK1-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
542 // CHECK1-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to %struct.TT**
543 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP112]], align 8
544 // CHECK1-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
545 // CHECK1-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to %struct.TT**
546 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP114]], align 8
547 // CHECK1-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 8
548 // CHECK1-NEXT:    store i8* null, i8** [[TMP115]], align 8
549 // CHECK1-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 9
550 // CHECK1-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i64*
551 // CHECK1-NEXT:    store i64 [[TMP63]], i64* [[TMP117]], align 8
552 // CHECK1-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 9
553 // CHECK1-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i64*
554 // CHECK1-NEXT:    store i64 [[TMP63]], i64* [[TMP119]], align 8
555 // CHECK1-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 9
556 // CHECK1-NEXT:    store i8* null, i8** [[TMP120]], align 8
557 // CHECK1-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
558 // CHECK1-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
559 // CHECK1-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
560 // CHECK1-NEXT:    [[TMP124:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140.region_id, i32 10, i8** [[TMP121]], i8** [[TMP122]], i64* [[TMP123]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
561 // CHECK1-NEXT:    [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
562 // CHECK1-NEXT:    br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
563 // CHECK1:       omp_offload.failed23:
564 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i64 [[TMP61]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP63]]) #[[ATTR4]]
565 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
566 // CHECK1:       omp_offload.cont24:
567 // CHECK1-NEXT:    br label [[OMP_IF_END26:%.*]]
568 // CHECK1:       omp_if.else25:
569 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i64 [[TMP61]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP63]]) #[[ATTR4]]
570 // CHECK1-NEXT:    br label [[OMP_IF_END26]]
571 // CHECK1:       omp_if.end26:
572 // CHECK1-NEXT:    [[TMP126:%.*]] = load i32, i32* [[A]], align 4
573 // CHECK1-NEXT:    [[TMP127:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
574 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP127]])
575 // CHECK1-NEXT:    ret i32 [[TMP126]]
576 //
577 //
578 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
579 // CHECK1-SAME: () #[[ATTR2:[0-9]+]] {
580 // CHECK1-NEXT:  entry:
581 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
582 // CHECK1-NEXT:    ret void
583 //
584 //
585 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
586 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
587 // CHECK1-NEXT:  entry:
588 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
589 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
590 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
591 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
592 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
593 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
594 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
595 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
596 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
597 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
598 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
599 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
600 // CHECK1-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
601 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
602 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
603 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
604 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
605 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
606 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
607 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
608 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
609 // CHECK1:       cond.true:
610 // CHECK1-NEXT:    br label [[COND_END:%.*]]
611 // CHECK1:       cond.false:
612 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
613 // CHECK1-NEXT:    br label [[COND_END]]
614 // CHECK1:       cond.end:
615 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
616 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
617 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
618 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
619 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
620 // CHECK1:       omp.inner.for.cond:
621 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
622 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
623 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
624 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
625 // CHECK1:       omp.inner.for.body:
626 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
627 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
628 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
629 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
630 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
631 // CHECK1:       omp.body.continue:
632 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
633 // CHECK1:       omp.inner.for.inc:
634 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
635 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
636 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
637 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
638 // CHECK1:       omp.inner.for.end:
639 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
640 // CHECK1:       omp.loop.exit:
641 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
642 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
643 // CHECK1-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
644 // CHECK1-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
645 // CHECK1:       .omp.final.then:
646 // CHECK1-NEXT:    store i32 33, i32* [[I]], align 4
647 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
648 // CHECK1:       .omp.final.done:
649 // CHECK1-NEXT:    ret void
650 //
651 //
652 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
653 // CHECK1-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
654 // CHECK1-NEXT:  entry:
655 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
656 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
657 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
658 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
659 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
660 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
661 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
662 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
663 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
664 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
665 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
666 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
667 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
668 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
669 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
670 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
671 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
672 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
673 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
674 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
675 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
676 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
677 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
678 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !25
679 // CHECK1-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !25
680 // CHECK1-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !25
681 // CHECK1-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !25
682 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !25
683 // CHECK1-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !25
684 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
685 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
686 // CHECK1-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
687 // CHECK1:       omp_offload.failed.i:
688 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96() #[[ATTR4]]
689 // CHECK1-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
690 // CHECK1:       .omp_outlined..1.exit:
691 // CHECK1-NEXT:    ret i32 0
692 //
693 //
694 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
695 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[K:%.*]]) #[[ATTR3]] {
696 // CHECK1-NEXT:  entry:
697 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
698 // CHECK1-NEXT:    [[K_ADDR:%.*]] = alloca i64, align 8
699 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
700 // CHECK1-NEXT:    [[K_CASTED:%.*]] = alloca i64, align 8
701 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
702 // CHECK1-NEXT:    store i64 [[K]], i64* [[K_ADDR]], align 8
703 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
704 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
705 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
706 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
707 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
708 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[K_ADDR]], align 8
709 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[K_CASTED]], align 8
710 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[K_CASTED]], align 8
711 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
712 // CHECK1-NEXT:    ret void
713 //
714 //
715 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
716 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[K:%.*]]) #[[ATTR3]] {
717 // CHECK1-NEXT:  entry:
718 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
719 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
720 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
721 // CHECK1-NEXT:    [[K_ADDR:%.*]] = alloca i64, align 8
722 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
723 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
724 // CHECK1-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
725 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
726 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
727 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
728 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
729 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
730 // CHECK1-NEXT:    [[K1:%.*]] = alloca i64, align 8
731 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
732 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
733 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
734 // CHECK1-NEXT:    store i64 [[K]], i64* [[K_ADDR]], align 8
735 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
736 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[K_ADDR]], align 8
737 // CHECK1-NEXT:    store i64 [[TMP0]], i64* [[DOTLINEAR_START]], align 8
738 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
739 // CHECK1-NEXT:    store i32 8, i32* [[DOTOMP_UB]], align 4
740 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
741 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
742 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
743 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
744 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]])
745 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 35, i32 0, i32 8, i32 1, i32 1)
746 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
747 // CHECK1:       omp.dispatch.cond:
748 // CHECK1-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
749 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP3]], 0
750 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
751 // CHECK1:       omp.dispatch.body:
752 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
753 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
754 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
755 // CHECK1:       omp.inner.for.cond:
756 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
757 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
758 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
759 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
760 // CHECK1:       omp.inner.for.body:
761 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
762 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
763 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL]]
764 // CHECK1-NEXT:    store i32 [[SUB]], i32* [[I]], align 4, !llvm.access.group !26
765 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !26
766 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
767 // CHECK1-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP9]], 3
768 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i32 [[MUL2]] to i64
769 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP8]], [[CONV3]]
770 // CHECK1-NEXT:    store i64 [[ADD]], i64* [[K1]], align 8, !llvm.access.group !26
771 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !26
772 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP10]], 1
773 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !26
774 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
775 // CHECK1:       omp.body.continue:
776 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
777 // CHECK1:       omp.inner.for.inc:
778 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
779 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
780 // CHECK1-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
781 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
782 // CHECK1:       omp.inner.for.end:
783 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
784 // CHECK1:       omp.dispatch.inc:
785 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
786 // CHECK1:       omp.dispatch.end:
787 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
788 // CHECK1-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
789 // CHECK1-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
790 // CHECK1:       .omp.final.then:
791 // CHECK1-NEXT:    store i32 1, i32* [[I]], align 4
792 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
793 // CHECK1:       .omp.final.done:
794 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
795 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
796 // CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
797 // CHECK1:       .omp.linear.pu:
798 // CHECK1-NEXT:    [[TMP16:%.*]] = load i64, i64* [[K1]], align 8
799 // CHECK1-NEXT:    store i64 [[TMP16]], i64* [[K_ADDR]], align 8
800 // CHECK1-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
801 // CHECK1:       .omp.linear.pu.done:
802 // CHECK1-NEXT:    ret void
803 //
804 //
805 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
806 // CHECK1-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
807 // CHECK1-NEXT:  entry:
808 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
809 // CHECK1-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
810 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
811 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
812 // CHECK1-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
813 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
814 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
815 // CHECK1-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
816 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
817 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
818 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
819 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
820 // CHECK1-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
821 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
822 // CHECK1-NEXT:    store i16 [[TMP0]], i16* [[CONV3]], align 2
823 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
824 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
825 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
826 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[CONV4]], align 4
827 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
828 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV2]], align 4
829 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
830 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
831 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[A_CASTED]], align 8
832 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]])
833 // CHECK1-NEXT:    ret void
834 //
835 //
836 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
837 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR3]] {
838 // CHECK1-NEXT:  entry:
839 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
840 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
841 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
842 // CHECK1-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
843 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
844 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
845 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
846 // CHECK1-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
847 // CHECK1-NEXT:    [[DOTLINEAR_START3:%.*]] = alloca i32, align 4
848 // CHECK1-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
849 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
850 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
851 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
852 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
853 // CHECK1-NEXT:    [[IT:%.*]] = alloca i64, align 8
854 // CHECK1-NEXT:    [[LIN4:%.*]] = alloca i32, align 4
855 // CHECK1-NEXT:    [[A5:%.*]] = alloca i32, align 4
856 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
857 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
858 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
859 // CHECK1-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
860 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
861 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
862 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
863 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
864 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV1]], align 4
865 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
866 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV2]], align 4
867 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START3]], align 4
868 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
869 // CHECK1-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
870 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
871 // CHECK1-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
872 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
873 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
874 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
875 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
876 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
877 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
878 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
879 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
880 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
881 // CHECK1:       cond.true:
882 // CHECK1-NEXT:    br label [[COND_END:%.*]]
883 // CHECK1:       cond.false:
884 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
885 // CHECK1-NEXT:    br label [[COND_END]]
886 // CHECK1:       cond.end:
887 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
888 // CHECK1-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
889 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
890 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
891 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
892 // CHECK1:       omp.inner.for.cond:
893 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
894 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !29
895 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
896 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
897 // CHECK1:       omp.inner.for.body:
898 // CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
899 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
900 // CHECK1-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
901 // CHECK1-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !29
902 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !29
903 // CHECK1-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP10]] to i64
904 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
905 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !29
906 // CHECK1-NEXT:    [[MUL8:%.*]] = mul i64 [[TMP11]], [[TMP12]]
907 // CHECK1-NEXT:    [[ADD:%.*]] = add i64 [[CONV7]], [[MUL8]]
908 // CHECK1-NEXT:    [[CONV9:%.*]] = trunc i64 [[ADD]] to i32
909 // CHECK1-NEXT:    store i32 [[CONV9]], i32* [[LIN4]], align 4, !llvm.access.group !29
910 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START3]], align 4, !llvm.access.group !29
911 // CHECK1-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
912 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
913 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !29
914 // CHECK1-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP14]], [[TMP15]]
915 // CHECK1-NEXT:    [[ADD12:%.*]] = add i64 [[CONV10]], [[MUL11]]
916 // CHECK1-NEXT:    [[CONV13:%.*]] = trunc i64 [[ADD12]] to i32
917 // CHECK1-NEXT:    store i32 [[CONV13]], i32* [[A5]], align 4, !llvm.access.group !29
918 // CHECK1-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !29
919 // CHECK1-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP16]] to i32
920 // CHECK1-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
921 // CHECK1-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
922 // CHECK1-NEXT:    store i16 [[CONV16]], i16* [[CONV]], align 2, !llvm.access.group !29
923 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
924 // CHECK1:       omp.body.continue:
925 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
926 // CHECK1:       omp.inner.for.inc:
927 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
928 // CHECK1-NEXT:    [[ADD17:%.*]] = add i64 [[TMP17]], 1
929 // CHECK1-NEXT:    store i64 [[ADD17]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
930 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
931 // CHECK1:       omp.inner.for.end:
932 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
933 // CHECK1:       omp.loop.exit:
934 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
935 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
936 // CHECK1-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
937 // CHECK1-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
938 // CHECK1:       .omp.final.then:
939 // CHECK1-NEXT:    store i64 400, i64* [[IT]], align 8
940 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
941 // CHECK1:       .omp.final.done:
942 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
943 // CHECK1-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
944 // CHECK1-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
945 // CHECK1:       .omp.linear.pu:
946 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN4]], align 4
947 // CHECK1-NEXT:    store i32 [[TMP22]], i32* [[CONV1]], align 4
948 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A5]], align 4
949 // CHECK1-NEXT:    store i32 [[TMP23]], i32* [[CONV2]], align 4
950 // CHECK1-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
951 // CHECK1:       .omp.linear.pu.done:
952 // CHECK1-NEXT:    ret void
953 //
954 //
955 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
956 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
957 // CHECK1-NEXT:  entry:
958 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
959 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
960 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
961 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
962 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
963 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
964 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
965 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
966 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
967 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
968 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
969 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
970 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
971 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
972 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
973 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
974 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
975 // CHECK1-NEXT:    ret void
976 //
977 //
978 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
979 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR3]] {
980 // CHECK1-NEXT:  entry:
981 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
982 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
983 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
984 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
985 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
986 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i16, align 2
987 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
988 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
989 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
990 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
991 // CHECK1-NEXT:    [[IT:%.*]] = alloca i16, align 2
992 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
993 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
994 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
995 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
996 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
997 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
998 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
999 // CHECK1-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
1000 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1001 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1002 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1003 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1004 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1005 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1006 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
1007 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1008 // CHECK1:       cond.true:
1009 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1010 // CHECK1:       cond.false:
1011 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1012 // CHECK1-NEXT:    br label [[COND_END]]
1013 // CHECK1:       cond.end:
1014 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1015 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1016 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1017 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1018 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1019 // CHECK1:       omp.inner.for.cond:
1020 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
1021 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
1022 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1023 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1024 // CHECK1:       omp.inner.for.body:
1025 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
1026 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
1027 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
1028 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i16
1029 // CHECK1-NEXT:    store i16 [[CONV3]], i16* [[IT]], align 2, !llvm.access.group !32
1030 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !32
1031 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
1032 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !32
1033 // CHECK1-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !32
1034 // CHECK1-NEXT:    [[CONV5:%.*]] = sext i16 [[TMP9]] to i32
1035 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
1036 // CHECK1-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i16
1037 // CHECK1-NEXT:    store i16 [[CONV7]], i16* [[CONV1]], align 2, !llvm.access.group !32
1038 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1039 // CHECK1:       omp.body.continue:
1040 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1041 // CHECK1:       omp.inner.for.inc:
1042 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
1043 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 1
1044 // CHECK1-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
1045 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
1046 // CHECK1:       omp.inner.for.end:
1047 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1048 // CHECK1:       omp.loop.exit:
1049 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1050 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1051 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1052 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1053 // CHECK1:       .omp.final.then:
1054 // CHECK1-NEXT:    store i16 22, i16* [[IT]], align 2
1055 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1056 // CHECK1:       .omp.final.done:
1057 // CHECK1-NEXT:    ret void
1058 //
1059 //
1060 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
1061 // CHECK1-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1062 // CHECK1-NEXT:  entry:
1063 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1064 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
1065 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1066 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
1067 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
1068 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1069 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
1070 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
1071 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
1072 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1073 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1074 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1075 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1076 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
1077 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1078 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
1079 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
1080 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1081 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
1082 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
1083 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
1084 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1085 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1086 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
1087 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1088 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
1089 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
1090 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1091 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
1092 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
1093 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
1094 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1095 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
1096 // CHECK1-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1097 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
1098 // CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
1099 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
1100 // CHECK1-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
1101 // CHECK1-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
1102 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
1103 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
1104 // CHECK1-NEXT:    ret void
1105 //
1106 //
1107 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
1108 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
1109 // CHECK1-NEXT:  entry:
1110 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1111 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1112 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1113 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
1114 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1115 // CHECK1-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
1116 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
1117 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1118 // CHECK1-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
1119 // CHECK1-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
1120 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
1121 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1122 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1123 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8, align 1
1124 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1125 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1126 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1127 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1128 // CHECK1-NEXT:    [[IT:%.*]] = alloca i8, align 1
1129 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1130 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1131 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1132 // CHECK1-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
1133 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1134 // CHECK1-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
1135 // CHECK1-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
1136 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1137 // CHECK1-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
1138 // CHECK1-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
1139 // CHECK1-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
1140 // CHECK1-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1141 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1142 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
1143 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1144 // CHECK1-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
1145 // CHECK1-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
1146 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1147 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
1148 // CHECK1-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
1149 // CHECK1-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
1150 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1151 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1152 // CHECK1-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
1153 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1154 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1155 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
1156 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1157 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1158 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
1159 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1160 // CHECK1:       omp.dispatch.cond:
1161 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1162 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
1163 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1164 // CHECK1:       cond.true:
1165 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1166 // CHECK1:       cond.false:
1167 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1168 // CHECK1-NEXT:    br label [[COND_END]]
1169 // CHECK1:       cond.end:
1170 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1171 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1172 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1173 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1174 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1175 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1176 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1177 // CHECK1-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1178 // CHECK1:       omp.dispatch.body:
1179 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1180 // CHECK1:       omp.inner.for.cond:
1181 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
1182 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !35
1183 // CHECK1-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1184 // CHECK1-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1185 // CHECK1:       omp.inner.for.body:
1186 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
1187 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1188 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
1189 // CHECK1-NEXT:    [[CONV8:%.*]] = trunc i32 [[SUB]] to i8
1190 // CHECK1-NEXT:    store i8 [[CONV8]], i8* [[IT]], align 1, !llvm.access.group !35
1191 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !35
1192 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
1193 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4, !llvm.access.group !35
1194 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
1195 // CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !35
1196 // CHECK1-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
1197 // CHECK1-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
1198 // CHECK1-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
1199 // CHECK1-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !35
1200 // CHECK1-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
1201 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !35
1202 // CHECK1-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
1203 // CHECK1-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
1204 // CHECK1-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
1205 // CHECK1-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !35
1206 // CHECK1-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
1207 // CHECK1-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
1208 // CHECK1-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !35
1209 // CHECK1-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
1210 // CHECK1-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !35
1211 // CHECK1-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
1212 // CHECK1-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
1213 // CHECK1-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
1214 // CHECK1-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !35
1215 // CHECK1-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
1216 // CHECK1-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !35
1217 // CHECK1-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
1218 // CHECK1-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !35
1219 // CHECK1-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
1220 // CHECK1-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !35
1221 // CHECK1-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
1222 // CHECK1-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !35
1223 // CHECK1-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
1224 // CHECK1-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
1225 // CHECK1-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
1226 // CHECK1-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !35
1227 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1228 // CHECK1:       omp.body.continue:
1229 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1230 // CHECK1:       omp.inner.for.inc:
1231 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
1232 // CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
1233 // CHECK1-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
1234 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
1235 // CHECK1:       omp.inner.for.end:
1236 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1237 // CHECK1:       omp.dispatch.inc:
1238 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1239 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1240 // CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
1241 // CHECK1-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
1242 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1243 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1244 // CHECK1-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
1245 // CHECK1-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
1246 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
1247 // CHECK1:       omp.dispatch.end:
1248 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
1249 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1250 // CHECK1-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
1251 // CHECK1-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1252 // CHECK1:       .omp.final.then:
1253 // CHECK1-NEXT:    store i8 96, i8* [[IT]], align 1
1254 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1255 // CHECK1:       .omp.final.done:
1256 // CHECK1-NEXT:    ret void
1257 //
1258 //
1259 // CHECK1-LABEL: define {{[^@]+}}@_Z3bari
1260 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1261 // CHECK1-NEXT:  entry:
1262 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1263 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1264 // CHECK1-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
1265 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1266 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1267 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1268 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
1269 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
1270 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
1271 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
1272 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
1273 // CHECK1-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
1274 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
1275 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
1276 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
1277 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1278 // CHECK1-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
1279 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
1280 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
1281 // CHECK1-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
1282 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1283 // CHECK1-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
1284 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
1285 // CHECK1-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
1286 // CHECK1-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
1287 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
1288 // CHECK1-NEXT:    ret i32 [[TMP8]]
1289 //
1290 //
1291 // CHECK1-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
1292 // CHECK1-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
1293 // CHECK1-NEXT:  entry:
1294 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1295 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1296 // CHECK1-NEXT:    [[B:%.*]] = alloca i32, align 4
1297 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1298 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1299 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1300 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
1301 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
1302 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
1303 // CHECK1-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
1304 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1305 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1306 // CHECK1-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1307 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1308 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
1309 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
1310 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1311 // CHECK1-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
1312 // CHECK1-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
1313 // CHECK1-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
1314 // CHECK1-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
1315 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
1316 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
1317 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
1318 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1319 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[CONV]], align 4
1320 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
1321 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
1322 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP7]], 60
1323 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1324 // CHECK1:       omp_if.then:
1325 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
1326 // CHECK1-NEXT:    [[TMP8:%.*]] = mul nuw i64 2, [[TMP2]]
1327 // CHECK1-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 2
1328 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1329 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes.11 to i8*), i64 40, i1 false)
1330 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1331 // CHECK1-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
1332 // CHECK1-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 8
1333 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1334 // CHECK1-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
1335 // CHECK1-NEXT:    store double* [[A]], double** [[TMP14]], align 8
1336 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1337 // CHECK1-NEXT:    store i8* null, i8** [[TMP15]], align 8
1338 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1339 // CHECK1-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
1340 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP17]], align 8
1341 // CHECK1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1342 // CHECK1-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
1343 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[TMP19]], align 8
1344 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1345 // CHECK1-NEXT:    store i8* null, i8** [[TMP20]], align 8
1346 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1347 // CHECK1-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
1348 // CHECK1-NEXT:    store i64 2, i64* [[TMP22]], align 8
1349 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1350 // CHECK1-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
1351 // CHECK1-NEXT:    store i64 2, i64* [[TMP24]], align 8
1352 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1353 // CHECK1-NEXT:    store i8* null, i8** [[TMP25]], align 8
1354 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1355 // CHECK1-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
1356 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP27]], align 8
1357 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1358 // CHECK1-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
1359 // CHECK1-NEXT:    store i64 [[TMP2]], i64* [[TMP29]], align 8
1360 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1361 // CHECK1-NEXT:    store i8* null, i8** [[TMP30]], align 8
1362 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1363 // CHECK1-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
1364 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 8
1365 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1366 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
1367 // CHECK1-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 8
1368 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1369 // CHECK1-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 8
1370 // CHECK1-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
1371 // CHECK1-NEXT:    store i8* null, i8** [[TMP36]], align 8
1372 // CHECK1-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1373 // CHECK1-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1374 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1375 // CHECK1-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.12, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1376 // CHECK1-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
1377 // CHECK1-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1378 // CHECK1:       omp_offload.failed:
1379 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
1380 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1381 // CHECK1:       omp_offload.cont:
1382 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1383 // CHECK1:       omp_if.else:
1384 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i64 [[TMP6]], i64 2, i64 [[TMP2]], i16* [[VLA]]) #[[ATTR4]]
1385 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1386 // CHECK1:       omp_if.end:
1387 // CHECK1-NEXT:    [[TMP42:%.*]] = mul nsw i64 1, [[TMP2]]
1388 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP42]]
1389 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1390 // CHECK1-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
1391 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP43]] to i32
1392 // CHECK1-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
1393 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], [[TMP44]]
1394 // CHECK1-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1395 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
1396 // CHECK1-NEXT:    ret i32 [[ADD4]]
1397 //
1398 //
1399 // CHECK1-LABEL: define {{[^@]+}}@_ZL7fstatici
1400 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
1401 // CHECK1-NEXT:  entry:
1402 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1403 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1404 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1405 // CHECK1-NEXT:    [[AAA:%.*]] = alloca i8, align 1
1406 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1407 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1408 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1409 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1410 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
1411 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
1412 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
1413 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1414 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1415 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1416 // CHECK1-NEXT:    store i8 0, i8* [[AAA]], align 1
1417 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1418 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1419 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1420 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1421 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1422 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1423 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1424 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1425 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
1426 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1427 // CHECK1-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
1428 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1429 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1430 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
1431 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1432 // CHECK1:       omp_if.then:
1433 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1434 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1435 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1436 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1437 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1438 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
1439 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1440 // CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
1441 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1442 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1443 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1444 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1445 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1446 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
1447 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1448 // CHECK1-NEXT:    store i8* null, i8** [[TMP16]], align 8
1449 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1450 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
1451 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
1452 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1453 // CHECK1-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
1454 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
1455 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1456 // CHECK1-NEXT:    store i8* null, i8** [[TMP21]], align 8
1457 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1458 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
1459 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
1460 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1461 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
1462 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
1463 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1464 // CHECK1-NEXT:    store i8* null, i8** [[TMP26]], align 8
1465 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1466 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1467 // CHECK1-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1468 // CHECK1-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
1469 // CHECK1-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1470 // CHECK1:       omp_offload.failed:
1471 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
1472 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1473 // CHECK1:       omp_offload.cont:
1474 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1475 // CHECK1:       omp_if.else:
1476 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
1477 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1478 // CHECK1:       omp_if.end:
1479 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
1480 // CHECK1-NEXT:    ret i32 [[TMP31]]
1481 //
1482 //
1483 // CHECK1-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
1484 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
1485 // CHECK1-NEXT:  entry:
1486 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1487 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
1488 // CHECK1-NEXT:    [[AA:%.*]] = alloca i16, align 2
1489 // CHECK1-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
1490 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1491 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1492 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1493 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1494 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1495 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1496 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
1497 // CHECK1-NEXT:    store i16 0, i16* [[AA]], align 2
1498 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
1499 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1500 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
1501 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
1502 // CHECK1-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
1503 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1504 // CHECK1-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
1505 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1506 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
1507 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
1508 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1509 // CHECK1:       omp_if.then:
1510 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1511 // CHECK1-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
1512 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
1513 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1514 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1515 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
1516 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1517 // CHECK1-NEXT:    store i8* null, i8** [[TMP9]], align 8
1518 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1519 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
1520 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
1521 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1522 // CHECK1-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1523 // CHECK1-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
1524 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1525 // CHECK1-NEXT:    store i8* null, i8** [[TMP14]], align 8
1526 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1527 // CHECK1-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
1528 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
1529 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1530 // CHECK1-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
1531 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
1532 // CHECK1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1533 // CHECK1-NEXT:    store i8* null, i8** [[TMP19]], align 8
1534 // CHECK1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1535 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1536 // CHECK1-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1537 // CHECK1-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
1538 // CHECK1-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1539 // CHECK1:       omp_offload.failed:
1540 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
1541 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1542 // CHECK1:       omp_offload.cont:
1543 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1544 // CHECK1:       omp_if.else:
1545 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
1546 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1547 // CHECK1:       omp_if.end:
1548 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
1549 // CHECK1-NEXT:    ret i32 [[TMP24]]
1550 //
1551 //
1552 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
1553 // CHECK1-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
1554 // CHECK1-NEXT:  entry:
1555 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1556 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1557 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1558 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1559 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1560 // CHECK1-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
1561 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1562 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1563 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1564 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1565 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1566 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1567 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1568 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1569 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1570 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1571 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
1572 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
1573 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
1574 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
1575 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], 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]])
1576 // CHECK1-NEXT:    ret void
1577 //
1578 //
1579 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
1580 // 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:%.*]]) #[[ATTR3]] {
1581 // CHECK1-NEXT:  entry:
1582 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1583 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1584 // CHECK1-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
1585 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
1586 // CHECK1-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
1587 // CHECK1-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
1588 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
1589 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1590 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
1591 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1592 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1593 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1594 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1595 // CHECK1-NEXT:    [[IT:%.*]] = alloca i64, align 8
1596 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1597 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1598 // CHECK1-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
1599 // CHECK1-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
1600 // CHECK1-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1601 // CHECK1-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
1602 // CHECK1-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
1603 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
1604 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
1605 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1606 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
1607 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
1608 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
1609 // CHECK1-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
1610 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1611 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1612 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1613 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
1614 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
1615 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1616 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP6]], 3
1617 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1618 // CHECK1:       cond.true:
1619 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1620 // CHECK1:       cond.false:
1621 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1622 // CHECK1-NEXT:    br label [[COND_END]]
1623 // CHECK1:       cond.end:
1624 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
1625 // CHECK1-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
1626 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1627 // CHECK1-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
1628 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1629 // CHECK1:       omp.inner.for.cond:
1630 // CHECK1-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
1631 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !38
1632 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP9]], [[TMP10]]
1633 // CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1634 // CHECK1:       omp.inner.for.body:
1635 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
1636 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP11]], 400
1637 // CHECK1-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
1638 // CHECK1-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !38
1639 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !38
1640 // CHECK1-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
1641 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double [[CONV4]], 1.500000e+00
1642 // CHECK1-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
1643 // CHECK1-NEXT:    store double [[ADD]], double* [[A]], align 8, !llvm.access.group !38
1644 // CHECK1-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
1645 // CHECK1-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 8, !llvm.access.group !38
1646 // CHECK1-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
1647 // CHECK1-NEXT:    store double [[INC]], double* [[A5]], align 8, !llvm.access.group !38
1648 // CHECK1-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
1649 // CHECK1-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
1650 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
1651 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
1652 // CHECK1-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2, !llvm.access.group !38
1653 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1654 // CHECK1:       omp.body.continue:
1655 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1656 // CHECK1:       omp.inner.for.inc:
1657 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
1658 // CHECK1-NEXT:    [[ADD8:%.*]] = add i64 [[TMP15]], 1
1659 // CHECK1-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
1660 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
1661 // CHECK1:       omp.inner.for.end:
1662 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1663 // CHECK1:       omp.loop.exit:
1664 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
1665 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1666 // CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1667 // CHECK1-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1668 // CHECK1:       .omp.final.then:
1669 // CHECK1-NEXT:    store i64 400, i64* [[IT]], align 8
1670 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1671 // CHECK1:       .omp.final.done:
1672 // CHECK1-NEXT:    ret void
1673 //
1674 //
1675 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
1676 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1677 // CHECK1-NEXT:  entry:
1678 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1679 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1680 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1681 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1682 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1683 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1684 // CHECK1-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
1685 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1686 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1687 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1688 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1689 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1690 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1691 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1692 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1693 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1694 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1695 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
1696 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1697 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
1698 // CHECK1-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1699 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
1700 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1701 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
1702 // CHECK1-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
1703 // CHECK1-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
1704 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
1705 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
1706 // CHECK1-NEXT:    ret void
1707 //
1708 //
1709 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
1710 // 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:%.*]]) #[[ATTR3]] {
1711 // CHECK1-NEXT:  entry:
1712 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1713 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1714 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1715 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1716 // CHECK1-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
1717 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1718 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1719 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1720 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1721 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1722 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1723 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1724 // CHECK1-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
1725 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1726 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1727 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1728 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
1729 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1730 // CHECK1-NEXT:    ret void
1731 //
1732 //
1733 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
1734 // CHECK1-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
1735 // CHECK1-NEXT:  entry:
1736 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1737 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1738 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1739 // CHECK1-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
1740 // CHECK1-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
1741 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1742 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1743 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1744 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1745 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1746 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1747 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
1748 // CHECK1-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
1749 // CHECK1-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
1750 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
1751 // CHECK1-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
1752 // CHECK1-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
1753 // CHECK1-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
1754 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
1755 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
1756 // CHECK1-NEXT:    ret void
1757 //
1758 //
1759 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..16
1760 // 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:%.*]]) #[[ATTR3]] {
1761 // CHECK1-NEXT:  entry:
1762 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1763 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1764 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
1765 // CHECK1-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
1766 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
1767 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1768 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
1769 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1770 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1771 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1772 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1773 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
1774 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1775 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1776 // CHECK1-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
1777 // CHECK1-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
1778 // CHECK1-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
1779 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
1780 // CHECK1-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
1781 // CHECK1-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
1782 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
1783 // CHECK1-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
1784 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1785 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1786 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1787 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1788 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
1789 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1790 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
1791 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1792 // CHECK1:       cond.true:
1793 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1794 // CHECK1:       cond.false:
1795 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1796 // CHECK1-NEXT:    br label [[COND_END]]
1797 // CHECK1:       cond.end:
1798 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1799 // CHECK1-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
1800 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1801 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
1802 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1803 // CHECK1:       omp.inner.for.cond:
1804 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !41
1805 // CHECK1-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !41
1806 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
1807 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1808 // CHECK1:       omp.inner.for.body:
1809 // CHECK1-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !41
1810 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
1811 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
1812 // CHECK1-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !41
1813 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !41
1814 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
1815 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4, !llvm.access.group !41
1816 // CHECK1-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !41
1817 // CHECK1-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
1818 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
1819 // CHECK1-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
1820 // CHECK1-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2, !llvm.access.group !41
1821 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
1822 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !41
1823 // CHECK1-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
1824 // CHECK1-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !41
1825 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1826 // CHECK1:       omp.body.continue:
1827 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1828 // CHECK1:       omp.inner.for.inc:
1829 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !41
1830 // CHECK1-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP12]], 1
1831 // CHECK1-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !41
1832 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
1833 // CHECK1:       omp.inner.for.end:
1834 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1835 // CHECK1:       omp.loop.exit:
1836 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1837 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1838 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1839 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1840 // CHECK1:       .omp.final.then:
1841 // CHECK1-NEXT:    store i64 11, i64* [[I]], align 8
1842 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1843 // CHECK1:       .omp.final.done:
1844 // CHECK1-NEXT:    ret void
1845 //
1846 //
1847 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1848 // CHECK1-SAME: () #[[ATTR8:[0-9]+]] {
1849 // CHECK1-NEXT:  entry:
1850 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1851 // CHECK1-NEXT:    ret void
1852 //
1853 //
1854 // CHECK3-LABEL: define {{[^@]+}}@_Z7get_valv
1855 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1856 // CHECK3-NEXT:  entry:
1857 // CHECK3-NEXT:    ret i64 0
1858 //
1859 //
1860 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooi
1861 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
1862 // CHECK3-NEXT:  entry:
1863 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1864 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
1865 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
1866 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
1867 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
1868 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1869 // CHECK3-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
1870 // CHECK3-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
1871 // CHECK3-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
1872 // CHECK3-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1873 // CHECK3-NEXT:    [[K:%.*]] = alloca i64, align 8
1874 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
1875 // CHECK3-NEXT:    [[LIN:%.*]] = alloca i32, align 4
1876 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
1877 // CHECK3-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
1878 // CHECK3-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
1879 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1880 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1881 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1882 // CHECK3-NEXT:    [[A_CASTED3:%.*]] = alloca i32, align 4
1883 // CHECK3-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
1884 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [2 x i8*], align 4
1885 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [2 x i8*], align 4
1886 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [2 x i8*], align 4
1887 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1888 // CHECK3-NEXT:    [[A_CASTED11:%.*]] = alloca i32, align 4
1889 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
1890 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS14:%.*]] = alloca [10 x i8*], align 4
1891 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS15:%.*]] = alloca [10 x i8*], align 4
1892 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS16:%.*]] = alloca [10 x i8*], align 4
1893 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
1894 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
1895 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1896 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
1897 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
1898 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1899 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
1900 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
1901 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
1902 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
1903 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
1904 // CHECK3-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
1905 // CHECK3-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
1906 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
1907 // CHECK3-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
1908 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
1909 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
1910 // CHECK3-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP5]])
1911 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
1912 // CHECK3-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
1913 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
1914 // CHECK3-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
1915 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
1916 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP10]], i64* [[K]]) #[[ATTR4:[0-9]+]]
1917 // CHECK3-NEXT:    store i32 12, i32* [[LIN]], align 4
1918 // CHECK3-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
1919 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
1920 // CHECK3-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
1921 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
1922 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[LIN]], align 4
1923 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[LIN_CASTED]], align 4
1924 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
1925 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A]], align 4
1926 // CHECK3-NEXT:    store i32 [[TMP15]], i32* [[A_CASTED2]], align 4
1927 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[A_CASTED2]], align 4
1928 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1929 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
1930 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP18]], align 4
1931 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1932 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
1933 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[TMP20]], align 4
1934 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1935 // CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
1936 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1937 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
1938 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[TMP23]], align 4
1939 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1940 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
1941 // CHECK3-NEXT:    store i32 [[TMP14]], i32* [[TMP25]], align 4
1942 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1943 // CHECK3-NEXT:    store i8* null, i8** [[TMP26]], align 4
1944 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1945 // CHECK3-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
1946 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[TMP28]], align 4
1947 // CHECK3-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1948 // CHECK3-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
1949 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[TMP30]], align 4
1950 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1951 // CHECK3-NEXT:    store i8* null, i8** [[TMP31]], align 4
1952 // CHECK3-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1953 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1954 // CHECK3-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108.region_id, i32 3, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1955 // CHECK3-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
1956 // CHECK3-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1957 // CHECK3:       omp_offload.failed:
1958 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108(i32 [[TMP12]], i32 [[TMP14]], i32 [[TMP16]]) #[[ATTR4]]
1959 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1960 // CHECK3:       omp_offload.cont:
1961 // CHECK3-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
1962 // CHECK3-NEXT:    store i32 [[TMP36]], i32* [[A_CASTED3]], align 4
1963 // CHECK3-NEXT:    [[TMP37:%.*]] = load i32, i32* [[A_CASTED3]], align 4
1964 // CHECK3-NEXT:    [[TMP38:%.*]] = load i16, i16* [[AA]], align 2
1965 // CHECK3-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
1966 // CHECK3-NEXT:    store i16 [[TMP38]], i16* [[CONV5]], align 2
1967 // CHECK3-NEXT:    [[TMP39:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
1968 // CHECK3-NEXT:    [[TMP40:%.*]] = load i32, i32* [[N_ADDR]], align 4
1969 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 10
1970 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1971 // CHECK3:       omp_if.then:
1972 // CHECK3-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
1973 // CHECK3-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
1974 // CHECK3-NEXT:    store i32 [[TMP37]], i32* [[TMP42]], align 4
1975 // CHECK3-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
1976 // CHECK3-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
1977 // CHECK3-NEXT:    store i32 [[TMP37]], i32* [[TMP44]], align 4
1978 // CHECK3-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
1979 // CHECK3-NEXT:    store i8* null, i8** [[TMP45]], align 4
1980 // CHECK3-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 1
1981 // CHECK3-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
1982 // CHECK3-NEXT:    store i32 [[TMP39]], i32* [[TMP47]], align 4
1983 // CHECK3-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 1
1984 // CHECK3-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
1985 // CHECK3-NEXT:    store i32 [[TMP39]], i32* [[TMP49]], align 4
1986 // CHECK3-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 1
1987 // CHECK3-NEXT:    store i8* null, i8** [[TMP50]], align 4
1988 // CHECK3-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
1989 // CHECK3-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
1990 // CHECK3-NEXT:    [[TMP53:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116.region_id, i32 2, i8** [[TMP51]], i8** [[TMP52]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
1991 // CHECK3-NEXT:    [[TMP54:%.*]] = icmp ne i32 [[TMP53]], 0
1992 // CHECK3-NEXT:    br i1 [[TMP54]], label [[OMP_OFFLOAD_FAILED9:%.*]], label [[OMP_OFFLOAD_CONT10:%.*]]
1993 // CHECK3:       omp_offload.failed9:
1994 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i32 [[TMP37]], i32 [[TMP39]]) #[[ATTR4]]
1995 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT10]]
1996 // CHECK3:       omp_offload.cont10:
1997 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
1998 // CHECK3:       omp_if.else:
1999 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i32 [[TMP37]], i32 [[TMP39]]) #[[ATTR4]]
2000 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2001 // CHECK3:       omp_if.end:
2002 // CHECK3-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
2003 // CHECK3-NEXT:    store i32 [[TMP55]], i32* [[DOTCAPTURE_EXPR_]], align 4
2004 // CHECK3-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A]], align 4
2005 // CHECK3-NEXT:    store i32 [[TMP56]], i32* [[A_CASTED11]], align 4
2006 // CHECK3-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A_CASTED11]], align 4
2007 // CHECK3-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2008 // CHECK3-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2009 // CHECK3-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2010 // CHECK3-NEXT:    [[TMP60:%.*]] = load i32, i32* [[N_ADDR]], align 4
2011 // CHECK3-NEXT:    [[CMP12:%.*]] = icmp sgt i32 [[TMP60]], 20
2012 // CHECK3-NEXT:    br i1 [[CMP12]], label [[OMP_IF_THEN13:%.*]], label [[OMP_IF_ELSE19:%.*]]
2013 // CHECK3:       omp_if.then13:
2014 // CHECK3-NEXT:    [[TMP61:%.*]] = mul nuw i32 [[TMP1]], 4
2015 // CHECK3-NEXT:    [[TMP62:%.*]] = sext i32 [[TMP61]] to i64
2016 // CHECK3-NEXT:    [[TMP63:%.*]] = mul nuw i32 5, [[TMP3]]
2017 // CHECK3-NEXT:    [[TMP64:%.*]] = mul nuw i32 [[TMP63]], 8
2018 // CHECK3-NEXT:    [[TMP65:%.*]] = sext i32 [[TMP64]] to i64
2019 // CHECK3-NEXT:    [[TMP66:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2020 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP66]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.8 to i8*), i32 80, i1 false)
2021 // CHECK3-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
2022 // CHECK3-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i32*
2023 // CHECK3-NEXT:    store i32 [[TMP57]], i32* [[TMP68]], align 4
2024 // CHECK3-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
2025 // CHECK3-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to i32*
2026 // CHECK3-NEXT:    store i32 [[TMP57]], i32* [[TMP70]], align 4
2027 // CHECK3-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 0
2028 // CHECK3-NEXT:    store i8* null, i8** [[TMP71]], align 4
2029 // CHECK3-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 1
2030 // CHECK3-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to [10 x float]**
2031 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP73]], align 4
2032 // CHECK3-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 1
2033 // CHECK3-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [10 x float]**
2034 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP75]], align 4
2035 // CHECK3-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 1
2036 // CHECK3-NEXT:    store i8* null, i8** [[TMP76]], align 4
2037 // CHECK3-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 2
2038 // CHECK3-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i32*
2039 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP78]], align 4
2040 // CHECK3-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 2
2041 // CHECK3-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
2042 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP80]], align 4
2043 // CHECK3-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 2
2044 // CHECK3-NEXT:    store i8* null, i8** [[TMP81]], align 4
2045 // CHECK3-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 3
2046 // CHECK3-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to float**
2047 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP83]], align 4
2048 // CHECK3-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 3
2049 // CHECK3-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to float**
2050 // CHECK3-NEXT:    store float* [[VLA]], float** [[TMP85]], align 4
2051 // CHECK3-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
2052 // CHECK3-NEXT:    store i64 [[TMP62]], i64* [[TMP86]], align 4
2053 // CHECK3-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 3
2054 // CHECK3-NEXT:    store i8* null, i8** [[TMP87]], align 4
2055 // CHECK3-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 4
2056 // CHECK3-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [5 x [10 x double]]**
2057 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP89]], align 4
2058 // CHECK3-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 4
2059 // CHECK3-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to [5 x [10 x double]]**
2060 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP91]], align 4
2061 // CHECK3-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 4
2062 // CHECK3-NEXT:    store i8* null, i8** [[TMP92]], align 4
2063 // CHECK3-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 5
2064 // CHECK3-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to i32*
2065 // CHECK3-NEXT:    store i32 5, i32* [[TMP94]], align 4
2066 // CHECK3-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 5
2067 // CHECK3-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i32*
2068 // CHECK3-NEXT:    store i32 5, i32* [[TMP96]], align 4
2069 // CHECK3-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 5
2070 // CHECK3-NEXT:    store i8* null, i8** [[TMP97]], align 4
2071 // CHECK3-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 6
2072 // CHECK3-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
2073 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP99]], align 4
2074 // CHECK3-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 6
2075 // CHECK3-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32*
2076 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP101]], align 4
2077 // CHECK3-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 6
2078 // CHECK3-NEXT:    store i8* null, i8** [[TMP102]], align 4
2079 // CHECK3-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 7
2080 // CHECK3-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to double**
2081 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP104]], align 4
2082 // CHECK3-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 7
2083 // CHECK3-NEXT:    [[TMP106:%.*]] = bitcast i8** [[TMP105]] to double**
2084 // CHECK3-NEXT:    store double* [[VLA1]], double** [[TMP106]], align 4
2085 // CHECK3-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
2086 // CHECK3-NEXT:    store i64 [[TMP65]], i64* [[TMP107]], align 4
2087 // CHECK3-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 7
2088 // CHECK3-NEXT:    store i8* null, i8** [[TMP108]], align 4
2089 // CHECK3-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 8
2090 // CHECK3-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to %struct.TT**
2091 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP110]], align 4
2092 // CHECK3-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 8
2093 // CHECK3-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to %struct.TT**
2094 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP112]], align 4
2095 // CHECK3-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 8
2096 // CHECK3-NEXT:    store i8* null, i8** [[TMP113]], align 4
2097 // CHECK3-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 9
2098 // CHECK3-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to i32*
2099 // CHECK3-NEXT:    store i32 [[TMP59]], i32* [[TMP115]], align 4
2100 // CHECK3-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 9
2101 // CHECK3-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32*
2102 // CHECK3-NEXT:    store i32 [[TMP59]], i32* [[TMP117]], align 4
2103 // CHECK3-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 9
2104 // CHECK3-NEXT:    store i8* null, i8** [[TMP118]], align 4
2105 // CHECK3-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
2106 // CHECK3-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
2107 // CHECK3-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2108 // CHECK3-NEXT:    [[TMP122:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140.region_id, i32 10, i8** [[TMP119]], i8** [[TMP120]], i64* [[TMP121]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2109 // CHECK3-NEXT:    [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
2110 // CHECK3-NEXT:    br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
2111 // CHECK3:       omp_offload.failed17:
2112 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i32 [[TMP57]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP59]]) #[[ATTR4]]
2113 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
2114 // CHECK3:       omp_offload.cont18:
2115 // CHECK3-NEXT:    br label [[OMP_IF_END20:%.*]]
2116 // CHECK3:       omp_if.else19:
2117 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i32 [[TMP57]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP59]]) #[[ATTR4]]
2118 // CHECK3-NEXT:    br label [[OMP_IF_END20]]
2119 // CHECK3:       omp_if.end20:
2120 // CHECK3-NEXT:    [[TMP124:%.*]] = load i32, i32* [[A]], align 4
2121 // CHECK3-NEXT:    [[TMP125:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2122 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP125]])
2123 // CHECK3-NEXT:    ret i32 [[TMP124]]
2124 //
2125 //
2126 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
2127 // CHECK3-SAME: () #[[ATTR2:[0-9]+]] {
2128 // CHECK3-NEXT:  entry:
2129 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
2130 // CHECK3-NEXT:    ret void
2131 //
2132 //
2133 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
2134 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
2135 // CHECK3-NEXT:  entry:
2136 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2137 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2138 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2139 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2140 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2141 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2142 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2143 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2144 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2145 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2146 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2147 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2148 // CHECK3-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
2149 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2150 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2151 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2152 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2153 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2154 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2155 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
2156 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2157 // CHECK3:       cond.true:
2158 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2159 // CHECK3:       cond.false:
2160 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2161 // CHECK3-NEXT:    br label [[COND_END]]
2162 // CHECK3:       cond.end:
2163 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2164 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2165 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2166 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2167 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2168 // CHECK3:       omp.inner.for.cond:
2169 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2170 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
2171 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2172 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2173 // CHECK3:       omp.inner.for.body:
2174 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2175 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
2176 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
2177 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
2178 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2179 // CHECK3:       omp.body.continue:
2180 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2181 // CHECK3:       omp.inner.for.inc:
2182 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2183 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
2184 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2185 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
2186 // CHECK3:       omp.inner.for.end:
2187 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2188 // CHECK3:       omp.loop.exit:
2189 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2190 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2191 // CHECK3-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
2192 // CHECK3-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2193 // CHECK3:       .omp.final.then:
2194 // CHECK3-NEXT:    store i32 33, i32* [[I]], align 4
2195 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2196 // CHECK3:       .omp.final.done:
2197 // CHECK3-NEXT:    ret void
2198 //
2199 //
2200 // CHECK3-LABEL: define {{[^@]+}}@.omp_task_entry.
2201 // CHECK3-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
2202 // CHECK3-NEXT:  entry:
2203 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
2204 // CHECK3-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
2205 // CHECK3-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
2206 // CHECK3-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
2207 // CHECK3-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
2208 // CHECK3-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
2209 // CHECK3-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
2210 // CHECK3-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
2211 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
2212 // CHECK3-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
2213 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
2214 // CHECK3-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
2215 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
2216 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
2217 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
2218 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
2219 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
2220 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
2221 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
2222 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
2223 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
2224 // CHECK3-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META24:![0-9]+]])
2225 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !26
2226 // CHECK3-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !26
2227 // CHECK3-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !26
2228 // CHECK3-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !26
2229 // CHECK3-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !26
2230 // CHECK3-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !26
2231 // CHECK3-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !26
2232 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
2233 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2234 // CHECK3-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
2235 // CHECK3:       omp_offload.failed.i:
2236 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96() #[[ATTR4]]
2237 // CHECK3-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
2238 // CHECK3:       .omp_outlined..1.exit:
2239 // CHECK3-NEXT:    ret i32 0
2240 //
2241 //
2242 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
2243 // CHECK3-SAME: (i32 noundef [[A:%.*]], i64* noundef nonnull align 4 dereferenceable(8) [[K:%.*]]) #[[ATTR3]] {
2244 // CHECK3-NEXT:  entry:
2245 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2246 // CHECK3-NEXT:    [[K_ADDR:%.*]] = alloca i64*, align 4
2247 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2248 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2249 // CHECK3-NEXT:    store i64* [[K]], i64** [[K_ADDR]], align 4
2250 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64*, i64** [[K_ADDR]], align 4
2251 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
2252 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
2253 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
2254 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i64*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], i64* [[TMP0]])
2255 // CHECK3-NEXT:    ret void
2256 //
2257 //
2258 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
2259 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i64* noundef nonnull align 4 dereferenceable(8) [[K:%.*]]) #[[ATTR3]] {
2260 // CHECK3-NEXT:  entry:
2261 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2262 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2263 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2264 // CHECK3-NEXT:    [[K_ADDR:%.*]] = alloca i64*, align 4
2265 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2266 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2267 // CHECK3-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
2268 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2269 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2270 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2271 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2272 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2273 // CHECK3-NEXT:    [[K1:%.*]] = alloca i64, align 8
2274 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2275 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2276 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2277 // CHECK3-NEXT:    store i64* [[K]], i64** [[K_ADDR]], align 4
2278 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64*, i64** [[K_ADDR]], align 4
2279 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[TMP0]], align 8
2280 // CHECK3-NEXT:    store i64 [[TMP1]], i64* [[DOTLINEAR_START]], align 8
2281 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2282 // CHECK3-NEXT:    store i32 8, i32* [[DOTOMP_UB]], align 4
2283 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2284 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2285 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2286 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2287 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
2288 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 35, i32 0, i32 8, i32 1, i32 1)
2289 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2290 // CHECK3:       omp.dispatch.cond:
2291 // CHECK3-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
2292 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP4]], 0
2293 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2294 // CHECK3:       omp.dispatch.body:
2295 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2296 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2297 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2298 // CHECK3:       omp.inner.for.cond:
2299 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2300 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
2301 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2302 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2303 // CHECK3:       omp.inner.for.body:
2304 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2305 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2306 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL]]
2307 // CHECK3-NEXT:    store i32 [[SUB]], i32* [[I]], align 4, !llvm.access.group !27
2308 // CHECK3-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !27
2309 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2310 // CHECK3-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP10]], 3
2311 // CHECK3-NEXT:    [[CONV:%.*]] = sext i32 [[MUL2]] to i64
2312 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP9]], [[CONV]]
2313 // CHECK3-NEXT:    store i64 [[ADD]], i64* [[K1]], align 8, !llvm.access.group !27
2314 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !27
2315 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
2316 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !27
2317 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2318 // CHECK3:       omp.body.continue:
2319 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2320 // CHECK3:       omp.inner.for.inc:
2321 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2322 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
2323 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2324 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
2325 // CHECK3:       omp.inner.for.end:
2326 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2327 // CHECK3:       omp.dispatch.inc:
2328 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
2329 // CHECK3:       omp.dispatch.end:
2330 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2331 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2332 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2333 // CHECK3:       .omp.final.then:
2334 // CHECK3-NEXT:    store i32 1, i32* [[I]], align 4
2335 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2336 // CHECK3:       .omp.final.done:
2337 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2338 // CHECK3-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
2339 // CHECK3-NEXT:    br i1 [[TMP16]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
2340 // CHECK3:       .omp.linear.pu:
2341 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[K1]], align 8
2342 // CHECK3-NEXT:    store i64 [[TMP17]], i64* [[TMP0]], align 8
2343 // CHECK3-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
2344 // CHECK3:       .omp.linear.pu.done:
2345 // CHECK3-NEXT:    ret void
2346 //
2347 //
2348 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
2349 // CHECK3-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
2350 // CHECK3-NEXT:  entry:
2351 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2352 // CHECK3-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
2353 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2354 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
2355 // CHECK3-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
2356 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2357 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2358 // CHECK3-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
2359 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2360 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2361 // CHECK3-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
2362 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
2363 // CHECK3-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
2364 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
2365 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
2366 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[LIN_CASTED]], align 4
2367 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
2368 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
2369 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[A_CASTED]], align 4
2370 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A_CASTED]], align 4
2371 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]])
2372 // CHECK3-NEXT:    ret void
2373 //
2374 //
2375 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
2376 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR3]] {
2377 // CHECK3-NEXT:  entry:
2378 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2379 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2380 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2381 // CHECK3-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
2382 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2383 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
2384 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 4
2385 // CHECK3-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
2386 // CHECK3-NEXT:    [[DOTLINEAR_START1:%.*]] = alloca i32, align 4
2387 // CHECK3-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
2388 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
2389 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
2390 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
2391 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2392 // CHECK3-NEXT:    [[IT:%.*]] = alloca i64, align 8
2393 // CHECK3-NEXT:    [[LIN2:%.*]] = alloca i32, align 4
2394 // CHECK3-NEXT:    [[A3:%.*]] = alloca i32, align 4
2395 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2396 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2397 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2398 // CHECK3-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
2399 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2400 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2401 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
2402 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
2403 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
2404 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START1]], align 4
2405 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
2406 // CHECK3-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
2407 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
2408 // CHECK3-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
2409 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
2410 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2411 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2412 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2413 // CHECK3-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
2414 // CHECK3-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
2415 // CHECK3-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2416 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
2417 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2418 // CHECK3:       cond.true:
2419 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2420 // CHECK3:       cond.false:
2421 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2422 // CHECK3-NEXT:    br label [[COND_END]]
2423 // CHECK3:       cond.end:
2424 // CHECK3-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2425 // CHECK3-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
2426 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
2427 // CHECK3-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
2428 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2429 // CHECK3:       omp.inner.for.cond:
2430 // CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2431 // CHECK3-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !30
2432 // CHECK3-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
2433 // CHECK3-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2434 // CHECK3:       omp.inner.for.body:
2435 // CHECK3-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2436 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
2437 // CHECK3-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
2438 // CHECK3-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !30
2439 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !30
2440 // CHECK3-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
2441 // CHECK3-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2442 // CHECK3-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !30
2443 // CHECK3-NEXT:    [[MUL6:%.*]] = mul i64 [[TMP11]], [[TMP12]]
2444 // CHECK3-NEXT:    [[ADD:%.*]] = add i64 [[CONV5]], [[MUL6]]
2445 // CHECK3-NEXT:    [[CONV7:%.*]] = trunc i64 [[ADD]] to i32
2446 // CHECK3-NEXT:    store i32 [[CONV7]], i32* [[LIN2]], align 4, !llvm.access.group !30
2447 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START1]], align 4, !llvm.access.group !30
2448 // CHECK3-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP13]] to i64
2449 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2450 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !30
2451 // CHECK3-NEXT:    [[MUL9:%.*]] = mul i64 [[TMP14]], [[TMP15]]
2452 // CHECK3-NEXT:    [[ADD10:%.*]] = add i64 [[CONV8]], [[MUL9]]
2453 // CHECK3-NEXT:    [[CONV11:%.*]] = trunc i64 [[ADD10]] to i32
2454 // CHECK3-NEXT:    store i32 [[CONV11]], i32* [[A3]], align 4, !llvm.access.group !30
2455 // CHECK3-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !30
2456 // CHECK3-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP16]] to i32
2457 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
2458 // CHECK3-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
2459 // CHECK3-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2, !llvm.access.group !30
2460 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2461 // CHECK3:       omp.body.continue:
2462 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2463 // CHECK3:       omp.inner.for.inc:
2464 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2465 // CHECK3-NEXT:    [[ADD15:%.*]] = add i64 [[TMP17]], 1
2466 // CHECK3-NEXT:    store i64 [[ADD15]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
2467 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
2468 // CHECK3:       omp.inner.for.end:
2469 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2470 // CHECK3:       omp.loop.exit:
2471 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2472 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2473 // CHECK3-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2474 // CHECK3-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2475 // CHECK3:       .omp.final.then:
2476 // CHECK3-NEXT:    store i64 400, i64* [[IT]], align 8
2477 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2478 // CHECK3:       .omp.final.done:
2479 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2480 // CHECK3-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
2481 // CHECK3-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
2482 // CHECK3:       .omp.linear.pu:
2483 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN2]], align 4
2484 // CHECK3-NEXT:    store i32 [[TMP22]], i32* [[LIN_ADDR]], align 4
2485 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A3]], align 4
2486 // CHECK3-NEXT:    store i32 [[TMP23]], i32* [[A_ADDR]], align 4
2487 // CHECK3-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
2488 // CHECK3:       .omp.linear.pu.done:
2489 // CHECK3-NEXT:    ret void
2490 //
2491 //
2492 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
2493 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
2494 // CHECK3-NEXT:  entry:
2495 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2496 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2497 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2498 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
2499 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2500 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2501 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2502 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
2503 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
2504 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
2505 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
2506 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
2507 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
2508 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
2509 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
2510 // CHECK3-NEXT:    ret void
2511 //
2512 //
2513 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
2514 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR3]] {
2515 // CHECK3-NEXT:  entry:
2516 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2517 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2518 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2519 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
2520 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2521 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i16, align 2
2522 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2523 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2524 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2525 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2526 // CHECK3-NEXT:    [[IT:%.*]] = alloca i16, align 2
2527 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2528 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2529 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2530 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
2531 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
2532 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2533 // CHECK3-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
2534 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2535 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2536 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2537 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2538 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2539 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2540 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
2541 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2542 // CHECK3:       cond.true:
2543 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2544 // CHECK3:       cond.false:
2545 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2546 // CHECK3-NEXT:    br label [[COND_END]]
2547 // CHECK3:       cond.end:
2548 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2549 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2550 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2551 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2552 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2553 // CHECK3:       omp.inner.for.cond:
2554 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
2555 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
2556 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2557 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2558 // CHECK3:       omp.inner.for.body:
2559 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
2560 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
2561 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
2562 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
2563 // CHECK3-NEXT:    store i16 [[CONV2]], i16* [[IT]], align 2, !llvm.access.group !33
2564 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !33
2565 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
2566 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !33
2567 // CHECK3-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !33
2568 // CHECK3-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
2569 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
2570 // CHECK3-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
2571 // CHECK3-NEXT:    store i16 [[CONV6]], i16* [[CONV]], align 2, !llvm.access.group !33
2572 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2573 // CHECK3:       omp.body.continue:
2574 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2575 // CHECK3:       omp.inner.for.inc:
2576 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
2577 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
2578 // CHECK3-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
2579 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
2580 // CHECK3:       omp.inner.for.end:
2581 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2582 // CHECK3:       omp.loop.exit:
2583 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2584 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2585 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2586 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2587 // CHECK3:       .omp.final.then:
2588 // CHECK3-NEXT:    store i16 22, i16* [[IT]], align 2
2589 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2590 // CHECK3:       .omp.final.done:
2591 // CHECK3-NEXT:    ret void
2592 //
2593 //
2594 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
2595 // CHECK3-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2596 // CHECK3-NEXT:  entry:
2597 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2598 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
2599 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2600 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
2601 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
2602 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
2603 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
2604 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
2605 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
2606 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2607 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2608 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2609 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2610 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
2611 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2612 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
2613 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
2614 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
2615 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
2616 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
2617 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
2618 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2619 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
2620 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2621 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
2622 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
2623 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
2624 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
2625 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
2626 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
2627 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
2628 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
2629 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
2630 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2631 // CHECK3-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2632 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2633 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
2634 // CHECK3-NEXT:    ret void
2635 //
2636 //
2637 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
2638 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
2639 // CHECK3-NEXT:  entry:
2640 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2641 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2642 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2643 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
2644 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
2645 // CHECK3-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
2646 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
2647 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
2648 // CHECK3-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
2649 // CHECK3-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
2650 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
2651 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2652 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2653 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i8, align 1
2654 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2655 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2656 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2657 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2658 // CHECK3-NEXT:    [[IT:%.*]] = alloca i8, align 1
2659 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2660 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2661 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
2662 // CHECK3-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
2663 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2664 // CHECK3-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
2665 // CHECK3-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
2666 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
2667 // CHECK3-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
2668 // CHECK3-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
2669 // CHECK3-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
2670 // CHECK3-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2671 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
2672 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2673 // CHECK3-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
2674 // CHECK3-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
2675 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
2676 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
2677 // CHECK3-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
2678 // CHECK3-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
2679 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2680 // CHECK3-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
2681 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2682 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2683 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2684 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2685 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2686 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
2687 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2688 // CHECK3:       omp.dispatch.cond:
2689 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2690 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
2691 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2692 // CHECK3:       cond.true:
2693 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2694 // CHECK3:       cond.false:
2695 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2696 // CHECK3-NEXT:    br label [[COND_END]]
2697 // CHECK3:       cond.end:
2698 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2699 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2700 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2701 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2702 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2703 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2704 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2705 // CHECK3-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2706 // CHECK3:       omp.dispatch.body:
2707 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2708 // CHECK3:       omp.inner.for.cond:
2709 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
2710 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !36
2711 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2712 // CHECK3-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2713 // CHECK3:       omp.inner.for.body:
2714 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
2715 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2716 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
2717 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[SUB]] to i8
2718 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[IT]], align 1, !llvm.access.group !36
2719 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !36
2720 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
2721 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4, !llvm.access.group !36
2722 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
2723 // CHECK3-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !36
2724 // CHECK3-NEXT:    [[CONV7:%.*]] = fpext float [[TMP20]] to double
2725 // CHECK3-NEXT:    [[ADD8:%.*]] = fadd double [[CONV7]], 1.000000e+00
2726 // CHECK3-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
2727 // CHECK3-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !36
2728 // CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
2729 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !36
2730 // CHECK3-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
2731 // CHECK3-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
2732 // CHECK3-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
2733 // CHECK3-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !36
2734 // CHECK3-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
2735 // CHECK3-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
2736 // CHECK3-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !36
2737 // CHECK3-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
2738 // CHECK3-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !36
2739 // CHECK3-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
2740 // CHECK3-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
2741 // CHECK3-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
2742 // CHECK3-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !36
2743 // CHECK3-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
2744 // CHECK3-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !36
2745 // CHECK3-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
2746 // CHECK3-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !36
2747 // CHECK3-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
2748 // CHECK3-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !36
2749 // CHECK3-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
2750 // CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !36
2751 // CHECK3-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
2752 // CHECK3-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
2753 // CHECK3-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
2754 // CHECK3-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !36
2755 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2756 // CHECK3:       omp.body.continue:
2757 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2758 // CHECK3:       omp.inner.for.inc:
2759 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
2760 // CHECK3-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
2761 // CHECK3-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
2762 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
2763 // CHECK3:       omp.inner.for.end:
2764 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2765 // CHECK3:       omp.dispatch.inc:
2766 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2767 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2768 // CHECK3-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
2769 // CHECK3-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
2770 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2771 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2772 // CHECK3-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
2773 // CHECK3-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
2774 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
2775 // CHECK3:       omp.dispatch.end:
2776 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
2777 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2778 // CHECK3-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
2779 // CHECK3-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2780 // CHECK3:       .omp.final.then:
2781 // CHECK3-NEXT:    store i8 96, i8* [[IT]], align 1
2782 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2783 // CHECK3:       .omp.final.done:
2784 // CHECK3-NEXT:    ret void
2785 //
2786 //
2787 // CHECK3-LABEL: define {{[^@]+}}@_Z3bari
2788 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
2789 // CHECK3-NEXT:  entry:
2790 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2791 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2792 // CHECK3-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
2793 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2794 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
2795 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2796 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
2797 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
2798 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
2799 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
2800 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2801 // CHECK3-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
2802 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
2803 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
2804 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
2805 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
2806 // CHECK3-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
2807 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
2808 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
2809 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
2810 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2811 // CHECK3-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
2812 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
2813 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
2814 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
2815 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
2816 // CHECK3-NEXT:    ret i32 [[TMP8]]
2817 //
2818 //
2819 // CHECK3-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
2820 // CHECK3-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
2821 // CHECK3-NEXT:  entry:
2822 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
2823 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2824 // CHECK3-NEXT:    [[B:%.*]] = alloca i32, align 4
2825 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
2826 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2827 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
2828 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
2829 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
2830 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
2831 // CHECK3-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
2832 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
2833 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2834 // CHECK3-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
2835 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2836 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
2837 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
2838 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
2839 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
2840 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
2841 // CHECK3-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
2842 // CHECK3-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
2843 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
2844 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B]], align 4
2845 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
2846 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
2847 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2848 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 60
2849 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2850 // CHECK3:       omp_if.then:
2851 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
2852 // CHECK3-NEXT:    [[TMP7:%.*]] = mul nuw i32 2, [[TMP1]]
2853 // CHECK3-NEXT:    [[TMP8:%.*]] = mul nuw i32 [[TMP7]], 2
2854 // CHECK3-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
2855 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2856 // CHECK3-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP10]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes.11 to i8*), i32 40, i1 false)
2857 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2858 // CHECK3-NEXT:    [[TMP12:%.*]] = bitcast i8** [[TMP11]] to %struct.S1**
2859 // CHECK3-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP12]], align 4
2860 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2861 // CHECK3-NEXT:    [[TMP14:%.*]] = bitcast i8** [[TMP13]] to double**
2862 // CHECK3-NEXT:    store double* [[A]], double** [[TMP14]], align 4
2863 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2864 // CHECK3-NEXT:    store i8* null, i8** [[TMP15]], align 4
2865 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2866 // CHECK3-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
2867 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP17]], align 4
2868 // CHECK3-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2869 // CHECK3-NEXT:    [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32*
2870 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP19]], align 4
2871 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2872 // CHECK3-NEXT:    store i8* null, i8** [[TMP20]], align 4
2873 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2874 // CHECK3-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
2875 // CHECK3-NEXT:    store i32 2, i32* [[TMP22]], align 4
2876 // CHECK3-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2877 // CHECK3-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i32*
2878 // CHECK3-NEXT:    store i32 2, i32* [[TMP24]], align 4
2879 // CHECK3-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2880 // CHECK3-NEXT:    store i8* null, i8** [[TMP25]], align 4
2881 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2882 // CHECK3-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
2883 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP27]], align 4
2884 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2885 // CHECK3-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i32*
2886 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP29]], align 4
2887 // CHECK3-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
2888 // CHECK3-NEXT:    store i8* null, i8** [[TMP30]], align 4
2889 // CHECK3-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
2890 // CHECK3-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i16**
2891 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP32]], align 4
2892 // CHECK3-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
2893 // CHECK3-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i16**
2894 // CHECK3-NEXT:    store i16* [[VLA]], i16** [[TMP34]], align 4
2895 // CHECK3-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
2896 // CHECK3-NEXT:    store i64 [[TMP9]], i64* [[TMP35]], align 4
2897 // CHECK3-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
2898 // CHECK3-NEXT:    store i8* null, i8** [[TMP36]], align 4
2899 // CHECK3-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2900 // CHECK3-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2901 // CHECK3-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2902 // CHECK3-NEXT:    [[TMP40:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216.region_id, i32 5, i8** [[TMP37]], i8** [[TMP38]], i64* [[TMP39]], i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes.12, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2903 // CHECK3-NEXT:    [[TMP41:%.*]] = icmp ne i32 [[TMP40]], 0
2904 // CHECK3-NEXT:    br i1 [[TMP41]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2905 // CHECK3:       omp_offload.failed:
2906 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
2907 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2908 // CHECK3:       omp_offload.cont:
2909 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
2910 // CHECK3:       omp_if.else:
2911 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216(%struct.S1* [[THIS1]], i32 [[TMP5]], i32 2, i32 [[TMP1]], i16* [[VLA]]) #[[ATTR4]]
2912 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2913 // CHECK3:       omp_if.end:
2914 // CHECK3-NEXT:    [[TMP42:%.*]] = mul nsw i32 1, [[TMP1]]
2915 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP42]]
2916 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
2917 // CHECK3-NEXT:    [[TMP43:%.*]] = load i16, i16* [[ARRAYIDX2]], align 2
2918 // CHECK3-NEXT:    [[CONV:%.*]] = sext i16 [[TMP43]] to i32
2919 // CHECK3-NEXT:    [[TMP44:%.*]] = load i32, i32* [[B]], align 4
2920 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[TMP44]]
2921 // CHECK3-NEXT:    [[TMP45:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2922 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP45]])
2923 // CHECK3-NEXT:    ret i32 [[ADD3]]
2924 //
2925 //
2926 // CHECK3-LABEL: define {{[^@]+}}@_ZL7fstatici
2927 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
2928 // CHECK3-NEXT:  entry:
2929 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2930 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
2931 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
2932 // CHECK3-NEXT:    [[AAA:%.*]] = alloca i8, align 1
2933 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
2934 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
2935 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
2936 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
2937 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
2938 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
2939 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
2940 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2941 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
2942 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
2943 // CHECK3-NEXT:    store i8 0, i8* [[AAA]], align 1
2944 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
2945 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
2946 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
2947 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
2948 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
2949 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
2950 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
2951 // CHECK3-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
2952 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
2953 // CHECK3-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
2954 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
2955 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2956 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
2957 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2958 // CHECK3:       omp_if.then:
2959 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2960 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
2961 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
2962 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2963 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
2964 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
2965 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2966 // CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
2967 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2968 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
2969 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
2970 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2971 // CHECK3-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
2972 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
2973 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2974 // CHECK3-NEXT:    store i8* null, i8** [[TMP16]], align 4
2975 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2976 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
2977 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
2978 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2979 // CHECK3-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
2980 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
2981 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2982 // CHECK3-NEXT:    store i8* null, i8** [[TMP21]], align 4
2983 // CHECK3-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2984 // CHECK3-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
2985 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
2986 // CHECK3-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2987 // CHECK3-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
2988 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
2989 // CHECK3-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
2990 // CHECK3-NEXT:    store i8* null, i8** [[TMP26]], align 4
2991 // CHECK3-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2992 // CHECK3-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2993 // CHECK3-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
2994 // CHECK3-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
2995 // CHECK3-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2996 // CHECK3:       omp_offload.failed:
2997 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
2998 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2999 // CHECK3:       omp_offload.cont:
3000 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3001 // CHECK3:       omp_if.else:
3002 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
3003 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3004 // CHECK3:       omp_if.end:
3005 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
3006 // CHECK3-NEXT:    ret i32 [[TMP31]]
3007 //
3008 //
3009 // CHECK3-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
3010 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
3011 // CHECK3-NEXT:  entry:
3012 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3013 // CHECK3-NEXT:    [[A:%.*]] = alloca i32, align 4
3014 // CHECK3-NEXT:    [[AA:%.*]] = alloca i16, align 2
3015 // CHECK3-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
3016 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3017 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3018 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3019 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3020 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3021 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3022 // CHECK3-NEXT:    store i32 0, i32* [[A]], align 4
3023 // CHECK3-NEXT:    store i16 0, i16* [[AA]], align 2
3024 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
3025 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
3026 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
3027 // CHECK3-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
3028 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3029 // CHECK3-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
3030 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3031 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3032 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
3033 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3034 // CHECK3:       omp_if.then:
3035 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3036 // CHECK3-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
3037 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
3038 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3039 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3040 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
3041 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3042 // CHECK3-NEXT:    store i8* null, i8** [[TMP9]], align 4
3043 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3044 // CHECK3-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
3045 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
3046 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3047 // CHECK3-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3048 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
3049 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3050 // CHECK3-NEXT:    store i8* null, i8** [[TMP14]], align 4
3051 // CHECK3-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3052 // CHECK3-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
3053 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
3054 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3055 // CHECK3-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
3056 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
3057 // CHECK3-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3058 // CHECK3-NEXT:    store i8* null, i8** [[TMP19]], align 4
3059 // CHECK3-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3060 // CHECK3-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3061 // CHECK3-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3062 // CHECK3-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
3063 // CHECK3-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3064 // CHECK3:       omp_offload.failed:
3065 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
3066 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3067 // CHECK3:       omp_offload.cont:
3068 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3069 // CHECK3:       omp_if.else:
3070 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
3071 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3072 // CHECK3:       omp_if.end:
3073 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
3074 // CHECK3-NEXT:    ret i32 [[TMP24]]
3075 //
3076 //
3077 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
3078 // CHECK3-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR2]] {
3079 // CHECK3-NEXT:  entry:
3080 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3081 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
3082 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3083 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3084 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
3085 // CHECK3-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
3086 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3087 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
3088 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3089 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3090 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
3091 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3092 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3093 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3094 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
3095 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
3096 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
3097 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
3098 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], 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]])
3099 // CHECK3-NEXT:    ret void
3100 //
3101 //
3102 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
3103 // 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:%.*]]) #[[ATTR3]] {
3104 // CHECK3-NEXT:  entry:
3105 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3106 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3107 // CHECK3-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
3108 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
3109 // CHECK3-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
3110 // CHECK3-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
3111 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
3112 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3113 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 4
3114 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3115 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3116 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3117 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3118 // CHECK3-NEXT:    [[IT:%.*]] = alloca i64, align 8
3119 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3120 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3121 // CHECK3-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
3122 // CHECK3-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
3123 // CHECK3-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3124 // CHECK3-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
3125 // CHECK3-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
3126 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
3127 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3128 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
3129 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
3130 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3131 // CHECK3-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
3132 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3133 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3134 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3135 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
3136 // CHECK3-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
3137 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3138 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP6]], 3
3139 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3140 // CHECK3:       cond.true:
3141 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3142 // CHECK3:       cond.false:
3143 // CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3144 // CHECK3-NEXT:    br label [[COND_END]]
3145 // CHECK3:       cond.end:
3146 // CHECK3-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
3147 // CHECK3-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
3148 // CHECK3-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3149 // CHECK3-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
3150 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3151 // CHECK3:       omp.inner.for.cond:
3152 // CHECK3-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
3153 // CHECK3-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !39
3154 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP9]], [[TMP10]]
3155 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3156 // CHECK3:       omp.inner.for.body:
3157 // CHECK3-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
3158 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP11]], 400
3159 // CHECK3-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
3160 // CHECK3-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !39
3161 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4, !llvm.access.group !39
3162 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
3163 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
3164 // CHECK3-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
3165 // CHECK3-NEXT:    store double [[ADD]], double* [[A]], align 4, !llvm.access.group !39
3166 // CHECK3-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
3167 // CHECK3-NEXT:    [[TMP13:%.*]] = load double, double* [[A4]], align 4, !llvm.access.group !39
3168 // CHECK3-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
3169 // CHECK3-NEXT:    store double [[INC]], double* [[A4]], align 4, !llvm.access.group !39
3170 // CHECK3-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
3171 // CHECK3-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
3172 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
3173 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
3174 // CHECK3-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2, !llvm.access.group !39
3175 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3176 // CHECK3:       omp.body.continue:
3177 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3178 // CHECK3:       omp.inner.for.inc:
3179 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
3180 // CHECK3-NEXT:    [[ADD7:%.*]] = add i64 [[TMP15]], 1
3181 // CHECK3-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
3182 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
3183 // CHECK3:       omp.inner.for.end:
3184 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3185 // CHECK3:       omp.loop.exit:
3186 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
3187 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3188 // CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
3189 // CHECK3-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3190 // CHECK3:       .omp.final.then:
3191 // CHECK3-NEXT:    store i64 400, i64* [[IT]], align 8
3192 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3193 // CHECK3:       .omp.final.done:
3194 // CHECK3-NEXT:    ret void
3195 //
3196 //
3197 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
3198 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3199 // CHECK3-NEXT:  entry:
3200 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3201 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3202 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
3203 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3204 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3205 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3206 // CHECK3-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
3207 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3208 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3209 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
3210 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3211 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3212 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
3213 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3214 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
3215 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
3216 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
3217 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
3218 // CHECK3-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3219 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
3220 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3221 // CHECK3-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
3222 // CHECK3-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
3223 // CHECK3-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
3224 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
3225 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
3226 // CHECK3-NEXT:    ret void
3227 //
3228 //
3229 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..13
3230 // 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:%.*]]) #[[ATTR3]] {
3231 // CHECK3-NEXT:  entry:
3232 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3233 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3234 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3235 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3236 // CHECK3-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
3237 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3238 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3239 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3240 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3241 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3242 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3243 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3244 // CHECK3-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
3245 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3246 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3247 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
3248 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3249 // CHECK3-NEXT:    ret void
3250 //
3251 //
3252 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
3253 // CHECK3-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
3254 // CHECK3-NEXT:  entry:
3255 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3256 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3257 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3258 // CHECK3-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
3259 // CHECK3-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
3260 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3261 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3262 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3263 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3264 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3265 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
3266 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
3267 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
3268 // CHECK3-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
3269 // CHECK3-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
3270 // CHECK3-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
3271 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
3272 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
3273 // CHECK3-NEXT:    ret void
3274 //
3275 //
3276 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..16
3277 // 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:%.*]]) #[[ATTR3]] {
3278 // CHECK3-NEXT:  entry:
3279 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3280 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3281 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
3282 // CHECK3-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
3283 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
3284 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3285 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 4
3286 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3287 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3288 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3289 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3290 // CHECK3-NEXT:    [[I:%.*]] = alloca i64, align 8
3291 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3292 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3293 // CHECK3-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
3294 // CHECK3-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
3295 // CHECK3-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
3296 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
3297 // CHECK3-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
3298 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3299 // CHECK3-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
3300 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3301 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3302 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3303 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3304 // CHECK3-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
3305 // CHECK3-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3306 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
3307 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3308 // CHECK3:       cond.true:
3309 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3310 // CHECK3:       cond.false:
3311 // CHECK3-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3312 // CHECK3-NEXT:    br label [[COND_END]]
3313 // CHECK3:       cond.end:
3314 // CHECK3-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3315 // CHECK3-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
3316 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3317 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
3318 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3319 // CHECK3:       omp.inner.for.cond:
3320 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !42
3321 // CHECK3-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !42
3322 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
3323 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3324 // CHECK3:       omp.inner.for.body:
3325 // CHECK3-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !42
3326 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
3327 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
3328 // CHECK3-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !42
3329 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !42
3330 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
3331 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4, !llvm.access.group !42
3332 // CHECK3-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !42
3333 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
3334 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
3335 // CHECK3-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
3336 // CHECK3-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2, !llvm.access.group !42
3337 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
3338 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !42
3339 // CHECK3-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
3340 // CHECK3-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !42
3341 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3342 // CHECK3:       omp.body.continue:
3343 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3344 // CHECK3:       omp.inner.for.inc:
3345 // CHECK3-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !42
3346 // CHECK3-NEXT:    [[ADD7:%.*]] = add nsw i64 [[TMP12]], 1
3347 // CHECK3-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !42
3348 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
3349 // CHECK3:       omp.inner.for.end:
3350 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3351 // CHECK3:       omp.loop.exit:
3352 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3353 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3354 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3355 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3356 // CHECK3:       .omp.final.then:
3357 // CHECK3-NEXT:    store i64 11, i64* [[I]], align 8
3358 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3359 // CHECK3:       .omp.final.done:
3360 // CHECK3-NEXT:    ret void
3361 //
3362 //
3363 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3364 // CHECK3-SAME: () #[[ATTR8:[0-9]+]] {
3365 // CHECK3-NEXT:  entry:
3366 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
3367 // CHECK3-NEXT:    ret void
3368 //
3369 //
3370 // CHECK5-LABEL: define {{[^@]+}}@_Z7get_valv
3371 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
3372 // CHECK5-NEXT:  entry:
3373 // CHECK5-NEXT:    ret i64 0
3374 //
3375 //
3376 // CHECK5-LABEL: define {{[^@]+}}@_Z3fooi
3377 // CHECK5-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
3378 // CHECK5-NEXT:  entry:
3379 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3380 // CHECK5-NEXT:    [[A:%.*]] = alloca i32, align 4
3381 // CHECK5-NEXT:    [[AA:%.*]] = alloca i16, align 2
3382 // CHECK5-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
3383 // CHECK5-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
3384 // CHECK5-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3385 // CHECK5-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
3386 // CHECK5-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
3387 // CHECK5-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
3388 // CHECK5-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3389 // CHECK5-NEXT:    [[K:%.*]] = alloca i64, align 8
3390 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3391 // CHECK5-NEXT:    [[K_CASTED:%.*]] = alloca i64, align 8
3392 // CHECK5-NEXT:    [[LIN:%.*]] = alloca i32, align 4
3393 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3394 // CHECK5-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
3395 // CHECK5-NEXT:    [[A_CASTED4:%.*]] = alloca i64, align 8
3396 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3397 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3398 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3399 // CHECK5-NEXT:    [[A_CASTED6:%.*]] = alloca i64, align 8
3400 // CHECK5-NEXT:    [[AA_CASTED8:%.*]] = alloca i64, align 8
3401 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [2 x i8*], align 8
3402 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS11:%.*]] = alloca [2 x i8*], align 8
3403 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [2 x i8*], align 8
3404 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3405 // CHECK5-NEXT:    [[A_CASTED15:%.*]] = alloca i64, align 8
3406 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3407 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [10 x i8*], align 8
3408 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS21:%.*]] = alloca [10 x i8*], align 8
3409 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [10 x i8*], align 8
3410 // CHECK5-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 8
3411 // CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
3412 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3413 // CHECK5-NEXT:    store i32 0, i32* [[A]], align 4
3414 // CHECK5-NEXT:    store i16 0, i16* [[AA]], align 2
3415 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
3416 // CHECK5-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
3417 // CHECK5-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
3418 // CHECK5-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
3419 // CHECK5-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP2]], align 4
3420 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
3421 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
3422 // CHECK5-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
3423 // CHECK5-NEXT:    [[TMP6:%.*]] = mul nuw i64 5, [[TMP5]]
3424 // CHECK5-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP6]], align 8
3425 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[__VLA_EXPR1]], align 8
3426 // CHECK5-NEXT:    [[TMP7:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i64 40, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
3427 // CHECK5-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.kmp_task_t_with_privates*
3428 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP8]], i32 0, i32 0
3429 // CHECK5-NEXT:    [[TMP10:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP7]])
3430 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
3431 // CHECK5-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
3432 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A]], align 4
3433 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3434 // CHECK5-NEXT:    store i32 [[TMP11]], i32* [[CONV]], align 4
3435 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[A_CASTED]], align 8
3436 // CHECK5-NEXT:    [[TMP13:%.*]] = load i64, i64* [[K]], align 8
3437 // CHECK5-NEXT:    store i64 [[TMP13]], i64* [[K_CASTED]], align 8
3438 // CHECK5-NEXT:    [[TMP14:%.*]] = load i64, i64* [[K_CASTED]], align 8
3439 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i64 [[TMP12]], i64 [[TMP14]]) #[[ATTR4:[0-9]+]]
3440 // CHECK5-NEXT:    store i32 12, i32* [[LIN]], align 4
3441 // CHECK5-NEXT:    [[TMP15:%.*]] = load i16, i16* [[AA]], align 2
3442 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3443 // CHECK5-NEXT:    store i16 [[TMP15]], i16* [[CONV2]], align 2
3444 // CHECK5-NEXT:    [[TMP16:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3445 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[LIN]], align 4
3446 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
3447 // CHECK5-NEXT:    store i32 [[TMP17]], i32* [[CONV3]], align 4
3448 // CHECK5-NEXT:    [[TMP18:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
3449 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A]], align 4
3450 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED4]] to i32*
3451 // CHECK5-NEXT:    store i32 [[TMP19]], i32* [[CONV5]], align 4
3452 // CHECK5-NEXT:    [[TMP20:%.*]] = load i64, i64* [[A_CASTED4]], align 8
3453 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3454 // CHECK5-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
3455 // CHECK5-NEXT:    store i64 [[TMP16]], i64* [[TMP22]], align 8
3456 // CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3457 // CHECK5-NEXT:    [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
3458 // CHECK5-NEXT:    store i64 [[TMP16]], i64* [[TMP24]], align 8
3459 // CHECK5-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3460 // CHECK5-NEXT:    store i8* null, i8** [[TMP25]], align 8
3461 // CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3462 // CHECK5-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
3463 // CHECK5-NEXT:    store i64 [[TMP18]], i64* [[TMP27]], align 8
3464 // CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3465 // CHECK5-NEXT:    [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
3466 // CHECK5-NEXT:    store i64 [[TMP18]], i64* [[TMP29]], align 8
3467 // CHECK5-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3468 // CHECK5-NEXT:    store i8* null, i8** [[TMP30]], align 8
3469 // CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3470 // CHECK5-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
3471 // CHECK5-NEXT:    store i64 [[TMP20]], i64* [[TMP32]], align 8
3472 // CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3473 // CHECK5-NEXT:    [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i64*
3474 // CHECK5-NEXT:    store i64 [[TMP20]], i64* [[TMP34]], align 8
3475 // CHECK5-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3476 // CHECK5-NEXT:    store i8* null, i8** [[TMP35]], align 8
3477 // CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3478 // CHECK5-NEXT:    [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3479 // CHECK5-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108.region_id, i32 3, i8** [[TMP36]], i8** [[TMP37]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3480 // CHECK5-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
3481 // CHECK5-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3482 // CHECK5:       omp_offload.failed:
3483 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108(i64 [[TMP16]], i64 [[TMP18]], i64 [[TMP20]]) #[[ATTR4]]
3484 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3485 // CHECK5:       omp_offload.cont:
3486 // CHECK5-NEXT:    [[TMP40:%.*]] = load i32, i32* [[A]], align 4
3487 // CHECK5-NEXT:    [[CONV7:%.*]] = bitcast i64* [[A_CASTED6]] to i32*
3488 // CHECK5-NEXT:    store i32 [[TMP40]], i32* [[CONV7]], align 4
3489 // CHECK5-NEXT:    [[TMP41:%.*]] = load i64, i64* [[A_CASTED6]], align 8
3490 // CHECK5-NEXT:    [[TMP42:%.*]] = load i16, i16* [[AA]], align 2
3491 // CHECK5-NEXT:    [[CONV9:%.*]] = bitcast i64* [[AA_CASTED8]] to i16*
3492 // CHECK5-NEXT:    store i16 [[TMP42]], i16* [[CONV9]], align 2
3493 // CHECK5-NEXT:    [[TMP43:%.*]] = load i64, i64* [[AA_CASTED8]], align 8
3494 // CHECK5-NEXT:    [[TMP44:%.*]] = load i32, i32* [[N_ADDR]], align 4
3495 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP44]], 10
3496 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3497 // CHECK5:       omp_if.then:
3498 // CHECK5-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3499 // CHECK5-NEXT:    [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
3500 // CHECK5-NEXT:    store i64 [[TMP41]], i64* [[TMP46]], align 8
3501 // CHECK5-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3502 // CHECK5-NEXT:    [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
3503 // CHECK5-NEXT:    store i64 [[TMP41]], i64* [[TMP48]], align 8
3504 // CHECK5-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
3505 // CHECK5-NEXT:    store i8* null, i8** [[TMP49]], align 8
3506 // CHECK5-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 1
3507 // CHECK5-NEXT:    [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
3508 // CHECK5-NEXT:    store i64 [[TMP43]], i64* [[TMP51]], align 8
3509 // CHECK5-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 1
3510 // CHECK5-NEXT:    [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
3511 // CHECK5-NEXT:    store i64 [[TMP43]], i64* [[TMP53]], align 8
3512 // CHECK5-NEXT:    [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 1
3513 // CHECK5-NEXT:    store i8* null, i8** [[TMP54]], align 8
3514 // CHECK5-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3515 // CHECK5-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3516 // CHECK5-NEXT:    [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3517 // CHECK5-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
3518 // CHECK5-NEXT:    br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
3519 // CHECK5:       omp_offload.failed13:
3520 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i64 [[TMP41]], i64 [[TMP43]]) #[[ATTR4]]
3521 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT14]]
3522 // CHECK5:       omp_offload.cont14:
3523 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
3524 // CHECK5:       omp_if.else:
3525 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i64 [[TMP41]], i64 [[TMP43]]) #[[ATTR4]]
3526 // CHECK5-NEXT:    br label [[OMP_IF_END]]
3527 // CHECK5:       omp_if.end:
3528 // CHECK5-NEXT:    [[TMP59:%.*]] = load i32, i32* [[A]], align 4
3529 // CHECK5-NEXT:    store i32 [[TMP59]], i32* [[DOTCAPTURE_EXPR_]], align 4
3530 // CHECK5-NEXT:    [[TMP60:%.*]] = load i32, i32* [[A]], align 4
3531 // CHECK5-NEXT:    [[CONV16:%.*]] = bitcast i64* [[A_CASTED15]] to i32*
3532 // CHECK5-NEXT:    store i32 [[TMP60]], i32* [[CONV16]], align 4
3533 // CHECK5-NEXT:    [[TMP61:%.*]] = load i64, i64* [[A_CASTED15]], align 8
3534 // CHECK5-NEXT:    [[TMP62:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3535 // CHECK5-NEXT:    [[CONV17:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3536 // CHECK5-NEXT:    store i32 [[TMP62]], i32* [[CONV17]], align 4
3537 // CHECK5-NEXT:    [[TMP63:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3538 // CHECK5-NEXT:    [[TMP64:%.*]] = load i32, i32* [[N_ADDR]], align 4
3539 // CHECK5-NEXT:    [[CMP18:%.*]] = icmp sgt i32 [[TMP64]], 20
3540 // CHECK5-NEXT:    br i1 [[CMP18]], label [[OMP_IF_THEN19:%.*]], label [[OMP_IF_ELSE25:%.*]]
3541 // CHECK5:       omp_if.then19:
3542 // CHECK5-NEXT:    [[TMP65:%.*]] = mul nuw i64 [[TMP2]], 4
3543 // CHECK5-NEXT:    [[TMP66:%.*]] = mul nuw i64 5, [[TMP5]]
3544 // CHECK5-NEXT:    [[TMP67:%.*]] = mul nuw i64 [[TMP66]], 8
3545 // CHECK5-NEXT:    [[TMP68:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3546 // CHECK5-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP68]], i8* align 8 bitcast ([10 x i64]* @.offload_sizes.8 to i8*), i64 80, i1 false)
3547 // CHECK5-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
3548 // CHECK5-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to i64*
3549 // CHECK5-NEXT:    store i64 [[TMP61]], i64* [[TMP70]], align 8
3550 // CHECK5-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
3551 // CHECK5-NEXT:    [[TMP72:%.*]] = bitcast i8** [[TMP71]] to i64*
3552 // CHECK5-NEXT:    store i64 [[TMP61]], i64* [[TMP72]], align 8
3553 // CHECK5-NEXT:    [[TMP73:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 0
3554 // CHECK5-NEXT:    store i8* null, i8** [[TMP73]], align 8
3555 // CHECK5-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
3556 // CHECK5-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [10 x float]**
3557 // CHECK5-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP75]], align 8
3558 // CHECK5-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
3559 // CHECK5-NEXT:    [[TMP77:%.*]] = bitcast i8** [[TMP76]] to [10 x float]**
3560 // CHECK5-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP77]], align 8
3561 // CHECK5-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 1
3562 // CHECK5-NEXT:    store i8* null, i8** [[TMP78]], align 8
3563 // CHECK5-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 2
3564 // CHECK5-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i64*
3565 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[TMP80]], align 8
3566 // CHECK5-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 2
3567 // CHECK5-NEXT:    [[TMP82:%.*]] = bitcast i8** [[TMP81]] to i64*
3568 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[TMP82]], align 8
3569 // CHECK5-NEXT:    [[TMP83:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 2
3570 // CHECK5-NEXT:    store i8* null, i8** [[TMP83]], align 8
3571 // CHECK5-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 3
3572 // CHECK5-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to float**
3573 // CHECK5-NEXT:    store float* [[VLA]], float** [[TMP85]], align 8
3574 // CHECK5-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 3
3575 // CHECK5-NEXT:    [[TMP87:%.*]] = bitcast i8** [[TMP86]] to float**
3576 // CHECK5-NEXT:    store float* [[VLA]], float** [[TMP87]], align 8
3577 // CHECK5-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
3578 // CHECK5-NEXT:    store i64 [[TMP65]], i64* [[TMP88]], align 8
3579 // CHECK5-NEXT:    [[TMP89:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 3
3580 // CHECK5-NEXT:    store i8* null, i8** [[TMP89]], align 8
3581 // CHECK5-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 4
3582 // CHECK5-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to [5 x [10 x double]]**
3583 // CHECK5-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP91]], align 8
3584 // CHECK5-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 4
3585 // CHECK5-NEXT:    [[TMP93:%.*]] = bitcast i8** [[TMP92]] to [5 x [10 x double]]**
3586 // CHECK5-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP93]], align 8
3587 // CHECK5-NEXT:    [[TMP94:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 4
3588 // CHECK5-NEXT:    store i8* null, i8** [[TMP94]], align 8
3589 // CHECK5-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 5
3590 // CHECK5-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
3591 // CHECK5-NEXT:    store i64 5, i64* [[TMP96]], align 8
3592 // CHECK5-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 5
3593 // CHECK5-NEXT:    [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
3594 // CHECK5-NEXT:    store i64 5, i64* [[TMP98]], align 8
3595 // CHECK5-NEXT:    [[TMP99:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 5
3596 // CHECK5-NEXT:    store i8* null, i8** [[TMP99]], align 8
3597 // CHECK5-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 6
3598 // CHECK5-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i64*
3599 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[TMP101]], align 8
3600 // CHECK5-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 6
3601 // CHECK5-NEXT:    [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i64*
3602 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[TMP103]], align 8
3603 // CHECK5-NEXT:    [[TMP104:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 6
3604 // CHECK5-NEXT:    store i8* null, i8** [[TMP104]], align 8
3605 // CHECK5-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 7
3606 // CHECK5-NEXT:    [[TMP106:%.*]] = bitcast i8** [[TMP105]] to double**
3607 // CHECK5-NEXT:    store double* [[VLA1]], double** [[TMP106]], align 8
3608 // CHECK5-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 7
3609 // CHECK5-NEXT:    [[TMP108:%.*]] = bitcast i8** [[TMP107]] to double**
3610 // CHECK5-NEXT:    store double* [[VLA1]], double** [[TMP108]], align 8
3611 // CHECK5-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
3612 // CHECK5-NEXT:    store i64 [[TMP67]], i64* [[TMP109]], align 8
3613 // CHECK5-NEXT:    [[TMP110:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 7
3614 // CHECK5-NEXT:    store i8* null, i8** [[TMP110]], align 8
3615 // CHECK5-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 8
3616 // CHECK5-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to %struct.TT**
3617 // CHECK5-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP112]], align 8
3618 // CHECK5-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 8
3619 // CHECK5-NEXT:    [[TMP114:%.*]] = bitcast i8** [[TMP113]] to %struct.TT**
3620 // CHECK5-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP114]], align 8
3621 // CHECK5-NEXT:    [[TMP115:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 8
3622 // CHECK5-NEXT:    store i8* null, i8** [[TMP115]], align 8
3623 // CHECK5-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 9
3624 // CHECK5-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i64*
3625 // CHECK5-NEXT:    store i64 [[TMP63]], i64* [[TMP117]], align 8
3626 // CHECK5-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 9
3627 // CHECK5-NEXT:    [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i64*
3628 // CHECK5-NEXT:    store i64 [[TMP63]], i64* [[TMP119]], align 8
3629 // CHECK5-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i64 0, i64 9
3630 // CHECK5-NEXT:    store i8* null, i8** [[TMP120]], align 8
3631 // CHECK5-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
3632 // CHECK5-NEXT:    [[TMP122:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
3633 // CHECK5-NEXT:    [[TMP123:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3634 // CHECK5-NEXT:    [[TMP124:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140.region_id, i32 10, i8** [[TMP121]], i8** [[TMP122]], i64* [[TMP123]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
3635 // CHECK5-NEXT:    [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
3636 // CHECK5-NEXT:    br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED23:%.*]], label [[OMP_OFFLOAD_CONT24:%.*]]
3637 // CHECK5:       omp_offload.failed23:
3638 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i64 [[TMP61]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP63]]) #[[ATTR4]]
3639 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT24]]
3640 // CHECK5:       omp_offload.cont24:
3641 // CHECK5-NEXT:    br label [[OMP_IF_END26:%.*]]
3642 // CHECK5:       omp_if.else25:
3643 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i64 [[TMP61]], [10 x float]* [[B]], i64 [[TMP2]], float* [[VLA]], [5 x [10 x double]]* [[C]], i64 5, i64 [[TMP5]], double* [[VLA1]], %struct.TT* [[D]], i64 [[TMP63]]) #[[ATTR4]]
3644 // CHECK5-NEXT:    br label [[OMP_IF_END26]]
3645 // CHECK5:       omp_if.end26:
3646 // CHECK5-NEXT:    [[TMP126:%.*]] = load i32, i32* [[A]], align 4
3647 // CHECK5-NEXT:    [[TMP127:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3648 // CHECK5-NEXT:    call void @llvm.stackrestore(i8* [[TMP127]])
3649 // CHECK5-NEXT:    ret i32 [[TMP126]]
3650 //
3651 //
3652 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
3653 // CHECK5-SAME: () #[[ATTR2:[0-9]+]] {
3654 // CHECK5-NEXT:  entry:
3655 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
3656 // CHECK5-NEXT:    ret void
3657 //
3658 //
3659 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
3660 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
3661 // CHECK5-NEXT:  entry:
3662 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3663 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3664 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3665 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3666 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3667 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3668 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3669 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3670 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
3671 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3672 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3673 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3674 // CHECK5-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
3675 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3676 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3677 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3678 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3679 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3680 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3681 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
3682 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3683 // CHECK5:       cond.true:
3684 // CHECK5-NEXT:    br label [[COND_END:%.*]]
3685 // CHECK5:       cond.false:
3686 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3687 // CHECK5-NEXT:    br label [[COND_END]]
3688 // CHECK5:       cond.end:
3689 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3690 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3691 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3692 // CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3693 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3694 // CHECK5:       omp.inner.for.cond:
3695 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3696 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
3697 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3698 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3699 // CHECK5:       omp.inner.for.body:
3700 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3701 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
3702 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
3703 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
3704 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3705 // CHECK5:       omp.body.continue:
3706 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3707 // CHECK5:       omp.inner.for.inc:
3708 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3709 // CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
3710 // CHECK5-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3711 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
3712 // CHECK5:       omp.inner.for.end:
3713 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3714 // CHECK5:       omp.loop.exit:
3715 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3716 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3717 // CHECK5-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
3718 // CHECK5-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3719 // CHECK5:       .omp.final.then:
3720 // CHECK5-NEXT:    store i32 33, i32* [[I]], align 4
3721 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3722 // CHECK5:       .omp.final.done:
3723 // CHECK5-NEXT:    ret void
3724 //
3725 //
3726 // CHECK5-LABEL: define {{[^@]+}}@.omp_task_entry.
3727 // CHECK5-SAME: (i32 noundef signext [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
3728 // CHECK5-NEXT:  entry:
3729 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
3730 // CHECK5-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
3731 // CHECK5-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
3732 // CHECK5-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
3733 // CHECK5-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
3734 // CHECK5-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
3735 // CHECK5-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
3736 // CHECK5-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
3737 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
3738 // CHECK5-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3739 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
3740 // CHECK5-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
3741 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
3742 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
3743 // CHECK5-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
3744 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
3745 // CHECK5-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
3746 // CHECK5-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
3747 // CHECK5-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META16:![0-9]+]])
3748 // CHECK5-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META19:![0-9]+]])
3749 // CHECK5-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META21:![0-9]+]])
3750 // CHECK5-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META23:![0-9]+]])
3751 // CHECK5-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !25
3752 // CHECK5-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !25
3753 // CHECK5-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !25
3754 // CHECK5-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !25
3755 // CHECK5-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !25
3756 // CHECK5-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !25
3757 // CHECK5-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !25
3758 // CHECK5-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
3759 // CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3760 // CHECK5-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
3761 // CHECK5:       omp_offload.failed.i:
3762 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96() #[[ATTR4]]
3763 // CHECK5-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
3764 // CHECK5:       .omp_outlined..1.exit:
3765 // CHECK5-NEXT:    ret i32 0
3766 //
3767 //
3768 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
3769 // CHECK5-SAME: (i64 noundef [[A:%.*]], i64 noundef [[K:%.*]]) #[[ATTR3]] {
3770 // CHECK5-NEXT:  entry:
3771 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3772 // CHECK5-NEXT:    [[K_ADDR:%.*]] = alloca i64, align 8
3773 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3774 // CHECK5-NEXT:    [[K_CASTED:%.*]] = alloca i64, align 8
3775 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3776 // CHECK5-NEXT:    store i64 [[K]], i64* [[K_ADDR]], align 8
3777 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3778 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
3779 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3780 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[CONV1]], align 4
3781 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
3782 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[K_ADDR]], align 8
3783 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[K_CASTED]], align 8
3784 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[K_CASTED]], align 8
3785 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
3786 // CHECK5-NEXT:    ret void
3787 //
3788 //
3789 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
3790 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[K:%.*]]) #[[ATTR3]] {
3791 // CHECK5-NEXT:  entry:
3792 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3793 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3794 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3795 // CHECK5-NEXT:    [[K_ADDR:%.*]] = alloca i64, align 8
3796 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3797 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3798 // CHECK5-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
3799 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3800 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3801 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3802 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3803 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
3804 // CHECK5-NEXT:    [[K1:%.*]] = alloca i64, align 8
3805 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3806 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3807 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3808 // CHECK5-NEXT:    store i64 [[K]], i64* [[K_ADDR]], align 8
3809 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3810 // CHECK5-NEXT:    [[TMP0:%.*]] = load i64, i64* [[K_ADDR]], align 8
3811 // CHECK5-NEXT:    store i64 [[TMP0]], i64* [[DOTLINEAR_START]], align 8
3812 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3813 // CHECK5-NEXT:    store i32 8, i32* [[DOTOMP_UB]], align 4
3814 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3815 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3816 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3817 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3818 // CHECK5-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP2]])
3819 // CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32 1073741859, i32 0, i32 8, i32 1, i32 1)
3820 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3821 // CHECK5:       omp.dispatch.cond:
3822 // CHECK5-NEXT:    [[TMP3:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
3823 // CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP3]], 0
3824 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3825 // CHECK5:       omp.dispatch.body:
3826 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3827 // CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3828 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3829 // CHECK5:       omp.inner.for.cond:
3830 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3831 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !26
3832 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3833 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3834 // CHECK5:       omp.inner.for.body:
3835 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3836 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 1
3837 // CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL]]
3838 // CHECK5-NEXT:    store i32 [[SUB]], i32* [[I]], align 4, !llvm.access.group !26
3839 // CHECK5-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !26
3840 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3841 // CHECK5-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP9]], 3
3842 // CHECK5-NEXT:    [[CONV3:%.*]] = sext i32 [[MUL2]] to i64
3843 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP8]], [[CONV3]]
3844 // CHECK5-NEXT:    store i64 [[ADD]], i64* [[K1]], align 8, !llvm.access.group !26
3845 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !26
3846 // CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP10]], 1
3847 // CHECK5-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !26
3848 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3849 // CHECK5:       omp.body.continue:
3850 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3851 // CHECK5:       omp.inner.for.inc:
3852 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3853 // CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
3854 // CHECK5-NEXT:    store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3855 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
3856 // CHECK5:       omp.inner.for.end:
3857 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3858 // CHECK5:       omp.dispatch.inc:
3859 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
3860 // CHECK5:       omp.dispatch.end:
3861 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3862 // CHECK5-NEXT:    [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
3863 // CHECK5-NEXT:    br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3864 // CHECK5:       .omp.final.then:
3865 // CHECK5-NEXT:    store i32 1, i32* [[I]], align 4
3866 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3867 // CHECK5:       .omp.final.done:
3868 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3869 // CHECK5-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
3870 // CHECK5-NEXT:    br i1 [[TMP15]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
3871 // CHECK5:       .omp.linear.pu:
3872 // CHECK5-NEXT:    [[TMP16:%.*]] = load i64, i64* [[K1]], align 8
3873 // CHECK5-NEXT:    store i64 [[TMP16]], i64* [[K_ADDR]], align 8
3874 // CHECK5-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
3875 // CHECK5:       .omp.linear.pu.done:
3876 // CHECK5-NEXT:    ret void
3877 //
3878 //
3879 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
3880 // CHECK5-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR2]] {
3881 // CHECK5-NEXT:  entry:
3882 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3883 // CHECK5-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
3884 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3885 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
3886 // CHECK5-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
3887 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
3888 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3889 // CHECK5-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
3890 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3891 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3892 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
3893 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3894 // CHECK5-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
3895 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
3896 // CHECK5-NEXT:    store i16 [[TMP0]], i16* [[CONV3]], align 2
3897 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
3898 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
3899 // CHECK5-NEXT:    [[CONV4:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
3900 // CHECK5-NEXT:    store i32 [[TMP2]], i32* [[CONV4]], align 4
3901 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
3902 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV2]], align 4
3903 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
3904 // CHECK5-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
3905 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[A_CASTED]], align 8
3906 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]])
3907 // CHECK5-NEXT:    ret void
3908 //
3909 //
3910 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
3911 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR3]] {
3912 // CHECK5-NEXT:  entry:
3913 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3914 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3915 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
3916 // CHECK5-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
3917 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
3918 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3919 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
3920 // CHECK5-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
3921 // CHECK5-NEXT:    [[DOTLINEAR_START3:%.*]] = alloca i32, align 4
3922 // CHECK5-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
3923 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3924 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3925 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3926 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3927 // CHECK5-NEXT:    [[IT:%.*]] = alloca i64, align 8
3928 // CHECK5-NEXT:    [[LIN4:%.*]] = alloca i32, align 4
3929 // CHECK5-NEXT:    [[A5:%.*]] = alloca i32, align 4
3930 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3931 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3932 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
3933 // CHECK5-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
3934 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
3935 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
3936 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
3937 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
3938 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV1]], align 4
3939 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
3940 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV2]], align 4
3941 // CHECK5-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START3]], align 4
3942 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
3943 // CHECK5-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
3944 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3945 // CHECK5-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
3946 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3947 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3948 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3949 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3950 // CHECK5-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
3951 // CHECK5-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
3952 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3953 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
3954 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3955 // CHECK5:       cond.true:
3956 // CHECK5-NEXT:    br label [[COND_END:%.*]]
3957 // CHECK5:       cond.false:
3958 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3959 // CHECK5-NEXT:    br label [[COND_END]]
3960 // CHECK5:       cond.end:
3961 // CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3962 // CHECK5-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
3963 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3964 // CHECK5-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
3965 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3966 // CHECK5:       omp.inner.for.cond:
3967 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
3968 // CHECK5-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !29
3969 // CHECK5-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
3970 // CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3971 // CHECK5:       omp.inner.for.body:
3972 // CHECK5-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
3973 // CHECK5-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
3974 // CHECK5-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
3975 // CHECK5-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !29
3976 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !29
3977 // CHECK5-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP10]] to i64
3978 // CHECK5-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
3979 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !29
3980 // CHECK5-NEXT:    [[MUL8:%.*]] = mul i64 [[TMP11]], [[TMP12]]
3981 // CHECK5-NEXT:    [[ADD:%.*]] = add i64 [[CONV7]], [[MUL8]]
3982 // CHECK5-NEXT:    [[CONV9:%.*]] = trunc i64 [[ADD]] to i32
3983 // CHECK5-NEXT:    store i32 [[CONV9]], i32* [[LIN4]], align 4, !llvm.access.group !29
3984 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START3]], align 4, !llvm.access.group !29
3985 // CHECK5-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
3986 // CHECK5-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
3987 // CHECK5-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !29
3988 // CHECK5-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP14]], [[TMP15]]
3989 // CHECK5-NEXT:    [[ADD12:%.*]] = add i64 [[CONV10]], [[MUL11]]
3990 // CHECK5-NEXT:    [[CONV13:%.*]] = trunc i64 [[ADD12]] to i32
3991 // CHECK5-NEXT:    store i32 [[CONV13]], i32* [[A5]], align 4, !llvm.access.group !29
3992 // CHECK5-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !29
3993 // CHECK5-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP16]] to i32
3994 // CHECK5-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
3995 // CHECK5-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
3996 // CHECK5-NEXT:    store i16 [[CONV16]], i16* [[CONV]], align 2, !llvm.access.group !29
3997 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3998 // CHECK5:       omp.body.continue:
3999 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4000 // CHECK5:       omp.inner.for.inc:
4001 // CHECK5-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
4002 // CHECK5-NEXT:    [[ADD17:%.*]] = add i64 [[TMP17]], 1
4003 // CHECK5-NEXT:    store i64 [[ADD17]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
4004 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
4005 // CHECK5:       omp.inner.for.end:
4006 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4007 // CHECK5:       omp.loop.exit:
4008 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4009 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4010 // CHECK5-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
4011 // CHECK5-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4012 // CHECK5:       .omp.final.then:
4013 // CHECK5-NEXT:    store i64 400, i64* [[IT]], align 8
4014 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4015 // CHECK5:       .omp.final.done:
4016 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4017 // CHECK5-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
4018 // CHECK5-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
4019 // CHECK5:       .omp.linear.pu:
4020 // CHECK5-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN4]], align 4
4021 // CHECK5-NEXT:    store i32 [[TMP22]], i32* [[CONV1]], align 4
4022 // CHECK5-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A5]], align 4
4023 // CHECK5-NEXT:    store i32 [[TMP23]], i32* [[CONV2]], align 4
4024 // CHECK5-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
4025 // CHECK5:       .omp.linear.pu.done:
4026 // CHECK5-NEXT:    ret void
4027 //
4028 //
4029 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
4030 // CHECK5-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR2]] {
4031 // CHECK5-NEXT:  entry:
4032 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4033 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4034 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4035 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
4036 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4037 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4038 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4039 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4040 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
4041 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4042 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
4043 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
4044 // CHECK5-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
4045 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
4046 // CHECK5-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
4047 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
4048 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
4049 // CHECK5-NEXT:    ret void
4050 //
4051 //
4052 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..4
4053 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR3]] {
4054 // CHECK5-NEXT:  entry:
4055 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4056 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4057 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4058 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4059 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4060 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i16, align 2
4061 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4062 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4063 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4064 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4065 // CHECK5-NEXT:    [[IT:%.*]] = alloca i16, align 2
4066 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4067 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4068 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4069 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4070 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4071 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4072 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4073 // CHECK5-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
4074 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4075 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4076 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4077 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4078 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4079 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4080 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
4081 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4082 // CHECK5:       cond.true:
4083 // CHECK5-NEXT:    br label [[COND_END:%.*]]
4084 // CHECK5:       cond.false:
4085 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4086 // CHECK5-NEXT:    br label [[COND_END]]
4087 // CHECK5:       cond.end:
4088 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4089 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4090 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4091 // CHECK5-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4092 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4093 // CHECK5:       omp.inner.for.cond:
4094 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
4095 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !32
4096 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4097 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4098 // CHECK5:       omp.inner.for.body:
4099 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
4100 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
4101 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
4102 // CHECK5-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i16
4103 // CHECK5-NEXT:    store i16 [[CONV3]], i16* [[IT]], align 2, !llvm.access.group !32
4104 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !32
4105 // CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
4106 // CHECK5-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !32
4107 // CHECK5-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !32
4108 // CHECK5-NEXT:    [[CONV5:%.*]] = sext i16 [[TMP9]] to i32
4109 // CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
4110 // CHECK5-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i16
4111 // CHECK5-NEXT:    store i16 [[CONV7]], i16* [[CONV1]], align 2, !llvm.access.group !32
4112 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4113 // CHECK5:       omp.body.continue:
4114 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4115 // CHECK5:       omp.inner.for.inc:
4116 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
4117 // CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 1
4118 // CHECK5-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
4119 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
4120 // CHECK5:       omp.inner.for.end:
4121 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4122 // CHECK5:       omp.loop.exit:
4123 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4124 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4125 // CHECK5-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4126 // CHECK5-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4127 // CHECK5:       .omp.final.then:
4128 // CHECK5-NEXT:    store i16 22, i16* [[IT]], align 2
4129 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4130 // CHECK5:       .omp.final.done:
4131 // CHECK5-NEXT:    ret void
4132 //
4133 //
4134 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
4135 // CHECK5-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4136 // CHECK5-NEXT:  entry:
4137 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4138 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
4139 // CHECK5-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4140 // CHECK5-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
4141 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
4142 // CHECK5-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4143 // CHECK5-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
4144 // CHECK5-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
4145 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
4146 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4147 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4148 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4149 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4150 // CHECK5-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
4151 // CHECK5-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4152 // CHECK5-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
4153 // CHECK5-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
4154 // CHECK5-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4155 // CHECK5-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
4156 // CHECK5-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
4157 // CHECK5-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
4158 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4159 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4160 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
4161 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4162 // CHECK5-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
4163 // CHECK5-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
4164 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4165 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
4166 // CHECK5-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
4167 // CHECK5-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
4168 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4169 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
4170 // CHECK5-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4171 // CHECK5-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
4172 // CHECK5-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
4173 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
4174 // CHECK5-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
4175 // CHECK5-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
4176 // CHECK5-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4177 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
4178 // CHECK5-NEXT:    ret void
4179 //
4180 //
4181 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
4182 // CHECK5-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
4183 // CHECK5-NEXT:  entry:
4184 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4185 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4186 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4187 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
4188 // CHECK5-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4189 // CHECK5-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
4190 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
4191 // CHECK5-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4192 // CHECK5-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
4193 // CHECK5-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
4194 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
4195 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4196 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4197 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i8, align 1
4198 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4199 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4200 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4201 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4202 // CHECK5-NEXT:    [[IT:%.*]] = alloca i8, align 1
4203 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4204 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4205 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4206 // CHECK5-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
4207 // CHECK5-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4208 // CHECK5-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
4209 // CHECK5-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
4210 // CHECK5-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4211 // CHECK5-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
4212 // CHECK5-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
4213 // CHECK5-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
4214 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4215 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4216 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
4217 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4218 // CHECK5-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
4219 // CHECK5-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
4220 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4221 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
4222 // CHECK5-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
4223 // CHECK5-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
4224 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4225 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4226 // CHECK5-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
4227 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4228 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4229 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
4230 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4231 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4232 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
4233 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4234 // CHECK5:       omp.dispatch.cond:
4235 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4236 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
4237 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4238 // CHECK5:       cond.true:
4239 // CHECK5-NEXT:    br label [[COND_END:%.*]]
4240 // CHECK5:       cond.false:
4241 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4242 // CHECK5-NEXT:    br label [[COND_END]]
4243 // CHECK5:       cond.end:
4244 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4245 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4246 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4247 // CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4248 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4249 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4250 // CHECK5-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
4251 // CHECK5-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4252 // CHECK5:       omp.dispatch.body:
4253 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4254 // CHECK5:       omp.inner.for.cond:
4255 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
4256 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !35
4257 // CHECK5-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
4258 // CHECK5-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4259 // CHECK5:       omp.inner.for.body:
4260 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
4261 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
4262 // CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
4263 // CHECK5-NEXT:    [[CONV8:%.*]] = trunc i32 [[SUB]] to i8
4264 // CHECK5-NEXT:    store i8 [[CONV8]], i8* [[IT]], align 1, !llvm.access.group !35
4265 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !35
4266 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
4267 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4, !llvm.access.group !35
4268 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
4269 // CHECK5-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !35
4270 // CHECK5-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
4271 // CHECK5-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
4272 // CHECK5-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
4273 // CHECK5-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !35
4274 // CHECK5-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
4275 // CHECK5-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !35
4276 // CHECK5-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
4277 // CHECK5-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
4278 // CHECK5-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
4279 // CHECK5-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !35
4280 // CHECK5-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
4281 // CHECK5-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
4282 // CHECK5-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !35
4283 // CHECK5-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
4284 // CHECK5-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !35
4285 // CHECK5-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
4286 // CHECK5-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
4287 // CHECK5-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
4288 // CHECK5-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !35
4289 // CHECK5-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
4290 // CHECK5-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !35
4291 // CHECK5-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
4292 // CHECK5-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !35
4293 // CHECK5-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
4294 // CHECK5-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !35
4295 // CHECK5-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
4296 // CHECK5-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !35
4297 // CHECK5-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
4298 // CHECK5-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
4299 // CHECK5-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
4300 // CHECK5-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !35
4301 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4302 // CHECK5:       omp.body.continue:
4303 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4304 // CHECK5:       omp.inner.for.inc:
4305 // CHECK5-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
4306 // CHECK5-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
4307 // CHECK5-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
4308 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
4309 // CHECK5:       omp.inner.for.end:
4310 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4311 // CHECK5:       omp.dispatch.inc:
4312 // CHECK5-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4313 // CHECK5-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4314 // CHECK5-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
4315 // CHECK5-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
4316 // CHECK5-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4317 // CHECK5-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4318 // CHECK5-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
4319 // CHECK5-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
4320 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]]
4321 // CHECK5:       omp.dispatch.end:
4322 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
4323 // CHECK5-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4324 // CHECK5-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
4325 // CHECK5-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4326 // CHECK5:       .omp.final.then:
4327 // CHECK5-NEXT:    store i8 96, i8* [[IT]], align 1
4328 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4329 // CHECK5:       .omp.final.done:
4330 // CHECK5-NEXT:    ret void
4331 //
4332 //
4333 // CHECK5-LABEL: define {{[^@]+}}@_Z3bari
4334 // CHECK5-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
4335 // CHECK5-NEXT:  entry:
4336 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4337 // CHECK5-NEXT:    [[A:%.*]] = alloca i32, align 4
4338 // CHECK5-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
4339 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4340 // CHECK5-NEXT:    store i32 0, i32* [[A]], align 4
4341 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4342 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
4343 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
4344 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
4345 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
4346 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
4347 // CHECK5-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
4348 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
4349 // CHECK5-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
4350 // CHECK5-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
4351 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4352 // CHECK5-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
4353 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
4354 // CHECK5-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
4355 // CHECK5-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
4356 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4357 // CHECK5-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
4358 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
4359 // CHECK5-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
4360 // CHECK5-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
4361 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
4362 // CHECK5-NEXT:    ret i32 [[TMP8]]
4363 //
4364 //
4365 // CHECK5-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
4366 // CHECK5-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
4367 // CHECK5-NEXT:  entry:
4368 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4369 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4370 // CHECK5-NEXT:    [[B:%.*]] = alloca i32, align 4
4371 // CHECK5-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
4372 // CHECK5-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
4373 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4374 // CHECK5-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
4375 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4376 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [6 x i8*], align 8
4377 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [6 x i8*], align 8
4378 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [6 x i8*], align 8
4379 // CHECK5-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [6 x i64], align 8
4380 // CHECK5-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4381 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4382 // CHECK5-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4383 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4384 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
4385 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
4386 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
4387 // CHECK5-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
4388 // CHECK5-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
4389 // CHECK5-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
4390 // CHECK5-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
4391 // CHECK5-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
4392 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
4393 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
4394 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 60
4395 // CHECK5-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
4396 // CHECK5-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
4397 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B]], align 4
4398 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_CASTED]] to i32*
4399 // CHECK5-NEXT:    store i32 [[TMP6]], i32* [[CONV]], align 4
4400 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[B_CASTED]], align 8
4401 // CHECK5-NEXT:    [[TMP8:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4402 // CHECK5-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP8]] to i1
4403 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
4404 // CHECK5-NEXT:    [[FROMBOOL3:%.*]] = zext i1 [[TOBOOL]] to i8
4405 // CHECK5-NEXT:    store i8 [[FROMBOOL3]], i8* [[CONV2]], align 1
4406 // CHECK5-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4407 // CHECK5-NEXT:    [[TMP10:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4408 // CHECK5-NEXT:    [[TOBOOL4:%.*]] = trunc i8 [[TMP10]] to i1
4409 // CHECK5-NEXT:    br i1 [[TOBOOL4]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4410 // CHECK5:       omp_if.then:
4411 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
4412 // CHECK5-NEXT:    [[TMP11:%.*]] = mul nuw i64 2, [[TMP2]]
4413 // CHECK5-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
4414 // CHECK5-NEXT:    [[TMP13:%.*]] = bitcast [6 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
4415 // CHECK5-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP13]], i8* align 8 bitcast ([6 x i64]* @.offload_sizes.11 to i8*), i64 48, i1 false)
4416 // CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4417 // CHECK5-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.S1**
4418 // CHECK5-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP15]], align 8
4419 // CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4420 // CHECK5-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to double**
4421 // CHECK5-NEXT:    store double* [[A]], double** [[TMP17]], align 8
4422 // CHECK5-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4423 // CHECK5-NEXT:    store i8* null, i8** [[TMP18]], align 8
4424 // CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4425 // CHECK5-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
4426 // CHECK5-NEXT:    store i64 [[TMP7]], i64* [[TMP20]], align 8
4427 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4428 // CHECK5-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
4429 // CHECK5-NEXT:    store i64 [[TMP7]], i64* [[TMP22]], align 8
4430 // CHECK5-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
4431 // CHECK5-NEXT:    store i8* null, i8** [[TMP23]], align 8
4432 // CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4433 // CHECK5-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
4434 // CHECK5-NEXT:    store i64 2, i64* [[TMP25]], align 8
4435 // CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4436 // CHECK5-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
4437 // CHECK5-NEXT:    store i64 2, i64* [[TMP27]], align 8
4438 // CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4439 // CHECK5-NEXT:    store i8* null, i8** [[TMP28]], align 8
4440 // CHECK5-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4441 // CHECK5-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
4442 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[TMP30]], align 8
4443 // CHECK5-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4444 // CHECK5-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
4445 // CHECK5-NEXT:    store i64 [[TMP2]], i64* [[TMP32]], align 8
4446 // CHECK5-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
4447 // CHECK5-NEXT:    store i8* null, i8** [[TMP33]], align 8
4448 // CHECK5-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
4449 // CHECK5-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
4450 // CHECK5-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 8
4451 // CHECK5-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
4452 // CHECK5-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
4453 // CHECK5-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 8
4454 // CHECK5-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [6 x i64], [6 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
4455 // CHECK5-NEXT:    store i64 [[TMP12]], i64* [[TMP38]], align 8
4456 // CHECK5-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
4457 // CHECK5-NEXT:    store i8* null, i8** [[TMP39]], align 8
4458 // CHECK5-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
4459 // CHECK5-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
4460 // CHECK5-NEXT:    store i64 [[TMP9]], i64* [[TMP41]], align 8
4461 // CHECK5-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
4462 // CHECK5-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
4463 // CHECK5-NEXT:    store i64 [[TMP9]], i64* [[TMP43]], align 8
4464 // CHECK5-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 5
4465 // CHECK5-NEXT:    store i8* null, i8** [[TMP44]], align 8
4466 // CHECK5-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4467 // CHECK5-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4468 // CHECK5-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [6 x i64], [6 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4469 // CHECK5-NEXT:    [[TMP48:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4470 // CHECK5-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP48]] to i1
4471 // CHECK5-NEXT:    [[TMP49:%.*]] = select i1 [[TOBOOL5]], i32 0, i32 1
4472 // CHECK5-NEXT:    [[TMP50:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214.region_id, i32 6, i8** [[TMP45]], i8** [[TMP46]], i64* [[TMP47]], i64* getelementptr inbounds ([6 x i64], [6 x i64]* @.offload_maptypes.12, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP49]])
4473 // CHECK5-NEXT:    [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
4474 // CHECK5-NEXT:    br i1 [[TMP51]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4475 // CHECK5:       omp_offload.failed:
4476 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214(%struct.S1* [[THIS1]], i64 [[TMP7]], i64 2, i64 [[TMP2]], i16* [[VLA]], i64 [[TMP9]]) #[[ATTR4]]
4477 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4478 // CHECK5:       omp_offload.cont:
4479 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
4480 // CHECK5:       omp_if.else:
4481 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214(%struct.S1* [[THIS1]], i64 [[TMP7]], i64 2, i64 [[TMP2]], i16* [[VLA]], i64 [[TMP9]]) #[[ATTR4]]
4482 // CHECK5-NEXT:    br label [[OMP_IF_END]]
4483 // CHECK5:       omp_if.end:
4484 // CHECK5-NEXT:    [[TMP52:%.*]] = mul nsw i64 1, [[TMP2]]
4485 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP52]]
4486 // CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
4487 // CHECK5-NEXT:    [[TMP53:%.*]] = load i16, i16* [[ARRAYIDX6]], align 2
4488 // CHECK5-NEXT:    [[CONV7:%.*]] = sext i16 [[TMP53]] to i32
4489 // CHECK5-NEXT:    [[TMP54:%.*]] = load i32, i32* [[B]], align 4
4490 // CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i32 [[CONV7]], [[TMP54]]
4491 // CHECK5-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4492 // CHECK5-NEXT:    call void @llvm.stackrestore(i8* [[TMP55]])
4493 // CHECK5-NEXT:    ret i32 [[ADD8]]
4494 //
4495 //
4496 // CHECK5-LABEL: define {{[^@]+}}@_ZL7fstatici
4497 // CHECK5-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
4498 // CHECK5-NEXT:  entry:
4499 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4500 // CHECK5-NEXT:    [[A:%.*]] = alloca i32, align 4
4501 // CHECK5-NEXT:    [[AA:%.*]] = alloca i16, align 2
4502 // CHECK5-NEXT:    [[AAA:%.*]] = alloca i8, align 1
4503 // CHECK5-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4504 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4505 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
4506 // CHECK5-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
4507 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
4508 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
4509 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
4510 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4511 // CHECK5-NEXT:    store i32 0, i32* [[A]], align 4
4512 // CHECK5-NEXT:    store i16 0, i16* [[AA]], align 2
4513 // CHECK5-NEXT:    store i8 0, i8* [[AAA]], align 1
4514 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4515 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4516 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
4517 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
4518 // CHECK5-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4519 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
4520 // CHECK5-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
4521 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
4522 // CHECK5-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
4523 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
4524 // CHECK5-NEXT:    store i8 [[TMP4]], i8* [[CONV2]], align 1
4525 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
4526 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4527 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
4528 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4529 // CHECK5:       omp_if.then:
4530 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4531 // CHECK5-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
4532 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
4533 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4534 // CHECK5-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
4535 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[TMP10]], align 8
4536 // CHECK5-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4537 // CHECK5-NEXT:    store i8* null, i8** [[TMP11]], align 8
4538 // CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4539 // CHECK5-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
4540 // CHECK5-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
4541 // CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4542 // CHECK5-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
4543 // CHECK5-NEXT:    store i64 [[TMP3]], i64* [[TMP15]], align 8
4544 // CHECK5-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
4545 // CHECK5-NEXT:    store i8* null, i8** [[TMP16]], align 8
4546 // CHECK5-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4547 // CHECK5-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
4548 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[TMP18]], align 8
4549 // CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4550 // CHECK5-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
4551 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[TMP20]], align 8
4552 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4553 // CHECK5-NEXT:    store i8* null, i8** [[TMP21]], align 8
4554 // CHECK5-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4555 // CHECK5-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
4556 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 8
4557 // CHECK5-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4558 // CHECK5-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
4559 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 8
4560 // CHECK5-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
4561 // CHECK5-NEXT:    store i8* null, i8** [[TMP26]], align 8
4562 // CHECK5-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4563 // CHECK5-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4564 // CHECK5-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
4565 // CHECK5-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
4566 // CHECK5-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4567 // CHECK5:       omp_offload.failed:
4568 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
4569 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4570 // CHECK5:       omp_offload.cont:
4571 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
4572 // CHECK5:       omp_if.else:
4573 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
4574 // CHECK5-NEXT:    br label [[OMP_IF_END]]
4575 // CHECK5:       omp_if.end:
4576 // CHECK5-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
4577 // CHECK5-NEXT:    ret i32 [[TMP31]]
4578 //
4579 //
4580 // CHECK5-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
4581 // CHECK5-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
4582 // CHECK5-NEXT:  entry:
4583 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4584 // CHECK5-NEXT:    [[A:%.*]] = alloca i32, align 4
4585 // CHECK5-NEXT:    [[AA:%.*]] = alloca i16, align 2
4586 // CHECK5-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
4587 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4588 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
4589 // CHECK5-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
4590 // CHECK5-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
4591 // CHECK5-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
4592 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4593 // CHECK5-NEXT:    store i32 0, i32* [[A]], align 4
4594 // CHECK5-NEXT:    store i16 0, i16* [[AA]], align 2
4595 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
4596 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4597 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
4598 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
4599 // CHECK5-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
4600 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
4601 // CHECK5-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
4602 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
4603 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
4604 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
4605 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4606 // CHECK5:       omp_if.then:
4607 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4608 // CHECK5-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i64*
4609 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[TMP6]], align 8
4610 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4611 // CHECK5-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
4612 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[TMP8]], align 8
4613 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4614 // CHECK5-NEXT:    store i8* null, i8** [[TMP9]], align 8
4615 // CHECK5-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4616 // CHECK5-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i64*
4617 // CHECK5-NEXT:    store i64 [[TMP3]], i64* [[TMP11]], align 8
4618 // CHECK5-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4619 // CHECK5-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
4620 // CHECK5-NEXT:    store i64 [[TMP3]], i64* [[TMP13]], align 8
4621 // CHECK5-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
4622 // CHECK5-NEXT:    store i8* null, i8** [[TMP14]], align 8
4623 // CHECK5-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4624 // CHECK5-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
4625 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 8
4626 // CHECK5-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4627 // CHECK5-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
4628 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 8
4629 // CHECK5-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4630 // CHECK5-NEXT:    store i8* null, i8** [[TMP19]], align 8
4631 // CHECK5-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4632 // CHECK5-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4633 // CHECK5-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
4634 // CHECK5-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
4635 // CHECK5-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4636 // CHECK5:       omp_offload.failed:
4637 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
4638 // CHECK5-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4639 // CHECK5:       omp_offload.cont:
4640 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
4641 // CHECK5:       omp_if.else:
4642 // CHECK5-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
4643 // CHECK5-NEXT:    br label [[OMP_IF_END]]
4644 // CHECK5:       omp_if.end:
4645 // CHECK5-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
4646 // CHECK5-NEXT:    ret i32 [[TMP24]]
4647 //
4648 //
4649 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214
4650 // CHECK5-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4651 // CHECK5-NEXT:  entry:
4652 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4653 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
4654 // CHECK5-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4655 // CHECK5-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4656 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
4657 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4658 // CHECK5-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
4659 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4660 // CHECK5-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
4661 // CHECK5-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4662 // CHECK5-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
4663 // CHECK5-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4664 // CHECK5-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
4665 // CHECK5-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4666 // CHECK5-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4667 // CHECK5-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
4668 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4669 // CHECK5-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4670 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
4671 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4672 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4673 // CHECK5-NEXT:    [[TMP4:%.*]] = load i16*, i16** [[C_ADDR]], align 8
4674 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
4675 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 4
4676 // CHECK5-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i32*
4677 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[CONV4]], align 4
4678 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
4679 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
4680 // CHECK5-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
4681 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
4682 // CHECK5-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
4683 // CHECK5-NEXT:    store i8 [[FROMBOOL]], i8* [[CONV5]], align 1
4684 // CHECK5-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4685 // CHECK5-NEXT:    [[TMP9:%.*]] = load i8, i8* [[CONV3]], align 1
4686 // CHECK5-NEXT:    [[TOBOOL6:%.*]] = trunc i8 [[TMP9]] to i1
4687 // CHECK5-NEXT:    br i1 [[TOBOOL6]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4688 // CHECK5:       omp_if.then:
4689 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*, i64)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP6]], i64 [[TMP2]], i64 [[TMP3]], i16* [[TMP4]], i64 [[TMP8]])
4690 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
4691 // CHECK5:       omp_if.else:
4692 // CHECK5-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
4693 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
4694 // CHECK5-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
4695 // CHECK5-NEXT:    call void @.omp_outlined..10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i64 [[TMP6]], i64 [[TMP2]], i64 [[TMP3]], i16* [[TMP4]], i64 [[TMP8]]) #[[ATTR4]]
4696 // CHECK5-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
4697 // CHECK5-NEXT:    br label [[OMP_IF_END]]
4698 // CHECK5:       omp_if.end:
4699 // CHECK5-NEXT:    ret void
4700 //
4701 //
4702 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
4703 // CHECK5-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:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
4704 // CHECK5-NEXT:  entry:
4705 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4706 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4707 // CHECK5-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
4708 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
4709 // CHECK5-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
4710 // CHECK5-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
4711 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
4712 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4713 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4714 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
4715 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4716 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4717 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4718 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4719 // CHECK5-NEXT:    [[IT:%.*]] = alloca i64, align 8
4720 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4721 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4722 // CHECK5-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
4723 // CHECK5-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
4724 // CHECK5-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4725 // CHECK5-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
4726 // CHECK5-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
4727 // CHECK5-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4728 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
4729 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
4730 // CHECK5-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4731 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
4732 // CHECK5-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
4733 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
4734 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
4735 // CHECK5-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
4736 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
4737 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4738 // CHECK5-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV3]], align 1
4739 // CHECK5-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
4740 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4741 // CHECK5:       omp_if.then:
4742 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4743 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
4744 // CHECK5-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
4745 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4746 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP7]], 3
4747 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4748 // CHECK5:       cond.true:
4749 // CHECK5-NEXT:    br label [[COND_END:%.*]]
4750 // CHECK5:       cond.false:
4751 // CHECK5-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4752 // CHECK5-NEXT:    br label [[COND_END]]
4753 // CHECK5:       cond.end:
4754 // CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
4755 // CHECK5-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
4756 // CHECK5-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4757 // CHECK5-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_IV]], align 8
4758 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4759 // CHECK5:       omp.inner.for.cond:
4760 // CHECK5-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
4761 // CHECK5-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !38
4762 // CHECK5-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP10]], [[TMP11]]
4763 // CHECK5-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4764 // CHECK5:       omp.inner.for.body:
4765 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
4766 // CHECK5-NEXT:    [[MUL:%.*]] = mul i64 [[TMP12]], 400
4767 // CHECK5-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
4768 // CHECK5-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !38
4769 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !38
4770 // CHECK5-NEXT:    [[CONV5:%.*]] = sitofp i32 [[TMP13]] to double
4771 // CHECK5-NEXT:    [[ADD:%.*]] = fadd double [[CONV5]], 1.500000e+00
4772 // CHECK5-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
4773 // CHECK5-NEXT:    store double [[ADD]], double* [[A]], align 8, !nontemporal !39, !llvm.access.group !38
4774 // CHECK5-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
4775 // CHECK5-NEXT:    [[TMP14:%.*]] = load double, double* [[A6]], align 8, !nontemporal !39, !llvm.access.group !38
4776 // CHECK5-NEXT:    [[INC:%.*]] = fadd double [[TMP14]], 1.000000e+00
4777 // CHECK5-NEXT:    store double [[INC]], double* [[A6]], align 8, !nontemporal !39, !llvm.access.group !38
4778 // CHECK5-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
4779 // CHECK5-NEXT:    [[TMP15:%.*]] = mul nsw i64 1, [[TMP2]]
4780 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP15]]
4781 // CHECK5-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
4782 // CHECK5-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2, !llvm.access.group !38
4783 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4784 // CHECK5:       omp.body.continue:
4785 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4786 // CHECK5:       omp.inner.for.inc:
4787 // CHECK5-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
4788 // CHECK5-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
4789 // CHECK5-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !38
4790 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
4791 // CHECK5:       omp.inner.for.end:
4792 // CHECK5-NEXT:    br label [[OMP_IF_END:%.*]]
4793 // CHECK5:       omp_if.else:
4794 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4795 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
4796 // CHECK5-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
4797 // CHECK5-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4798 // CHECK5-NEXT:    [[CMP10:%.*]] = icmp ugt i64 [[TMP19]], 3
4799 // CHECK5-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
4800 // CHECK5:       cond.true11:
4801 // CHECK5-NEXT:    br label [[COND_END13:%.*]]
4802 // CHECK5:       cond.false12:
4803 // CHECK5-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4804 // CHECK5-NEXT:    br label [[COND_END13]]
4805 // CHECK5:       cond.end13:
4806 // CHECK5-NEXT:    [[COND14:%.*]] = phi i64 [ 3, [[COND_TRUE11]] ], [ [[TMP20]], [[COND_FALSE12]] ]
4807 // CHECK5-NEXT:    store i64 [[COND14]], i64* [[DOTOMP_UB]], align 8
4808 // CHECK5-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4809 // CHECK5-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV]], align 8
4810 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND15:%.*]]
4811 // CHECK5:       omp.inner.for.cond15:
4812 // CHECK5-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
4813 // CHECK5-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4814 // CHECK5-NEXT:    [[CMP16:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
4815 // CHECK5-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY17:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
4816 // CHECK5:       omp.inner.for.body17:
4817 // CHECK5-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
4818 // CHECK5-NEXT:    [[MUL18:%.*]] = mul i64 [[TMP24]], 400
4819 // CHECK5-NEXT:    [[SUB19:%.*]] = sub i64 2000, [[MUL18]]
4820 // CHECK5-NEXT:    store i64 [[SUB19]], i64* [[IT]], align 8
4821 // CHECK5-NEXT:    [[TMP25:%.*]] = load i32, i32* [[CONV]], align 4
4822 // CHECK5-NEXT:    [[CONV20:%.*]] = sitofp i32 [[TMP25]] to double
4823 // CHECK5-NEXT:    [[ADD21:%.*]] = fadd double [[CONV20]], 1.500000e+00
4824 // CHECK5-NEXT:    [[A22:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
4825 // CHECK5-NEXT:    store double [[ADD21]], double* [[A22]], align 8
4826 // CHECK5-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
4827 // CHECK5-NEXT:    [[TMP26:%.*]] = load double, double* [[A23]], align 8
4828 // CHECK5-NEXT:    [[INC24:%.*]] = fadd double [[TMP26]], 1.000000e+00
4829 // CHECK5-NEXT:    store double [[INC24]], double* [[A23]], align 8
4830 // CHECK5-NEXT:    [[CONV25:%.*]] = fptosi double [[INC24]] to i16
4831 // CHECK5-NEXT:    [[TMP27:%.*]] = mul nsw i64 1, [[TMP2]]
4832 // CHECK5-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP27]]
4833 // CHECK5-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX26]], i64 1
4834 // CHECK5-NEXT:    store i16 [[CONV25]], i16* [[ARRAYIDX27]], align 2
4835 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE28:%.*]]
4836 // CHECK5:       omp.body.continue28:
4837 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC29:%.*]]
4838 // CHECK5:       omp.inner.for.inc29:
4839 // CHECK5-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
4840 // CHECK5-NEXT:    [[ADD30:%.*]] = add i64 [[TMP28]], 1
4841 // CHECK5-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
4842 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND15]], !llvm.loop [[LOOP42:![0-9]+]]
4843 // CHECK5:       omp.inner.for.end31:
4844 // CHECK5-NEXT:    br label [[OMP_IF_END]]
4845 // CHECK5:       omp_if.end:
4846 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4847 // CHECK5:       omp.loop.exit:
4848 // CHECK5-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4849 // CHECK5-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
4850 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
4851 // CHECK5-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4852 // CHECK5-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
4853 // CHECK5-NEXT:    br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4854 // CHECK5:       .omp.final.then:
4855 // CHECK5-NEXT:    store i64 400, i64* [[IT]], align 8
4856 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4857 // CHECK5:       .omp.final.done:
4858 // CHECK5-NEXT:    ret void
4859 //
4860 //
4861 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
4862 // CHECK5-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4863 // CHECK5-NEXT:  entry:
4864 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4865 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4866 // CHECK5-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
4867 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4868 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4869 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
4870 // CHECK5-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
4871 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4872 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4873 // CHECK5-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
4874 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4875 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4876 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4877 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
4878 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4879 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
4880 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4881 // CHECK5-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
4882 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
4883 // CHECK5-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
4884 // CHECK5-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
4885 // CHECK5-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
4886 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
4887 // CHECK5-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
4888 // CHECK5-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
4889 // CHECK5-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
4890 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
4891 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
4892 // CHECK5-NEXT:    ret void
4893 //
4894 //
4895 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..13
4896 // CHECK5-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:%.*]]) #[[ATTR3]] {
4897 // CHECK5-NEXT:  entry:
4898 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4899 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4900 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4901 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4902 // CHECK5-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
4903 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4904 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4905 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4906 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4907 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4908 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4909 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4910 // CHECK5-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
4911 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4912 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4913 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4914 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
4915 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4916 // CHECK5-NEXT:    ret void
4917 //
4918 //
4919 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
4920 // CHECK5-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
4921 // CHECK5-NEXT:  entry:
4922 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4923 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4924 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4925 // CHECK5-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
4926 // CHECK5-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
4927 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4928 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4929 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4930 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4931 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4932 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4933 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
4934 // CHECK5-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
4935 // CHECK5-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
4936 // CHECK5-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
4937 // CHECK5-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
4938 // CHECK5-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
4939 // CHECK5-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
4940 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
4941 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
4942 // CHECK5-NEXT:    ret void
4943 //
4944 //
4945 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..16
4946 // CHECK5-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:%.*]]) #[[ATTR3]] {
4947 // CHECK5-NEXT:  entry:
4948 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4949 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4950 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
4951 // CHECK5-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
4952 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
4953 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4954 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
4955 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4956 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4957 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4958 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4959 // CHECK5-NEXT:    [[I:%.*]] = alloca i64, align 8
4960 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4961 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4962 // CHECK5-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
4963 // CHECK5-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
4964 // CHECK5-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
4965 // CHECK5-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
4966 // CHECK5-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
4967 // CHECK5-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
4968 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
4969 // CHECK5-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
4970 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
4971 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4972 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4973 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4974 // CHECK5-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
4975 // CHECK5-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4976 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
4977 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4978 // CHECK5:       cond.true:
4979 // CHECK5-NEXT:    br label [[COND_END:%.*]]
4980 // CHECK5:       cond.false:
4981 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
4982 // CHECK5-NEXT:    br label [[COND_END]]
4983 // CHECK5:       cond.end:
4984 // CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4985 // CHECK5-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
4986 // CHECK5-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4987 // CHECK5-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
4988 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4989 // CHECK5:       omp.inner.for.cond:
4990 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !44
4991 // CHECK5-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !44
4992 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
4993 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4994 // CHECK5:       omp.inner.for.body:
4995 // CHECK5-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !44
4996 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
4997 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
4998 // CHECK5-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !44
4999 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !44
5000 // CHECK5-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
5001 // CHECK5-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4, !llvm.access.group !44
5002 // CHECK5-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !44
5003 // CHECK5-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
5004 // CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
5005 // CHECK5-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
5006 // CHECK5-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2, !llvm.access.group !44
5007 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
5008 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
5009 // CHECK5-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
5010 // CHECK5-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !44
5011 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5012 // CHECK5:       omp.body.continue:
5013 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5014 // CHECK5:       omp.inner.for.inc:
5015 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !44
5016 // CHECK5-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP12]], 1
5017 // CHECK5-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !44
5018 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
5019 // CHECK5:       omp.inner.for.end:
5020 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5021 // CHECK5:       omp.loop.exit:
5022 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5023 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5024 // CHECK5-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5025 // CHECK5-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5026 // CHECK5:       .omp.final.then:
5027 // CHECK5-NEXT:    store i64 11, i64* [[I]], align 8
5028 // CHECK5-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5029 // CHECK5:       .omp.final.done:
5030 // CHECK5-NEXT:    ret void
5031 //
5032 //
5033 // CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5034 // CHECK5-SAME: () #[[ATTR8:[0-9]+]] {
5035 // CHECK5-NEXT:  entry:
5036 // CHECK5-NEXT:    call void @__tgt_register_requires(i64 1)
5037 // CHECK5-NEXT:    ret void
5038 //
5039 //
5040 // CHECK7-LABEL: define {{[^@]+}}@_Z7get_valv
5041 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
5042 // CHECK7-NEXT:  entry:
5043 // CHECK7-NEXT:    ret i64 0
5044 //
5045 //
5046 // CHECK7-LABEL: define {{[^@]+}}@_Z3fooi
5047 // CHECK7-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
5048 // CHECK7-NEXT:  entry:
5049 // CHECK7-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5050 // CHECK7-NEXT:    [[A:%.*]] = alloca i32, align 4
5051 // CHECK7-NEXT:    [[AA:%.*]] = alloca i16, align 2
5052 // CHECK7-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
5053 // CHECK7-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
5054 // CHECK7-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5055 // CHECK7-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
5056 // CHECK7-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
5057 // CHECK7-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
5058 // CHECK7-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
5059 // CHECK7-NEXT:    [[K:%.*]] = alloca i64, align 8
5060 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5061 // CHECK7-NEXT:    [[LIN:%.*]] = alloca i32, align 4
5062 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5063 // CHECK7-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
5064 // CHECK7-NEXT:    [[A_CASTED2:%.*]] = alloca i32, align 4
5065 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
5066 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
5067 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
5068 // CHECK7-NEXT:    [[A_CASTED3:%.*]] = alloca i32, align 4
5069 // CHECK7-NEXT:    [[AA_CASTED4:%.*]] = alloca i32, align 4
5070 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS6:%.*]] = alloca [2 x i8*], align 4
5071 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS7:%.*]] = alloca [2 x i8*], align 4
5072 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS8:%.*]] = alloca [2 x i8*], align 4
5073 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5074 // CHECK7-NEXT:    [[A_CASTED11:%.*]] = alloca i32, align 4
5075 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
5076 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS14:%.*]] = alloca [10 x i8*], align 4
5077 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS15:%.*]] = alloca [10 x i8*], align 4
5078 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS16:%.*]] = alloca [10 x i8*], align 4
5079 // CHECK7-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [10 x i64], align 4
5080 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
5081 // CHECK7-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5082 // CHECK7-NEXT:    store i32 0, i32* [[A]], align 4
5083 // CHECK7-NEXT:    store i16 0, i16* [[AA]], align 2
5084 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
5085 // CHECK7-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
5086 // CHECK7-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
5087 // CHECK7-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP1]], align 4
5088 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
5089 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
5090 // CHECK7-NEXT:    [[TMP4:%.*]] = mul nuw i32 5, [[TMP3]]
5091 // CHECK7-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP4]], align 8
5092 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[__VLA_EXPR1]], align 4
5093 // CHECK7-NEXT:    [[TMP5:%.*]] = call i8* @__kmpc_omp_target_task_alloc(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 1, i32 20, i32 1, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*), i64 -1)
5094 // CHECK7-NEXT:    [[TMP6:%.*]] = bitcast i8* [[TMP5]] to %struct.kmp_task_t_with_privates*
5095 // CHECK7-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP6]], i32 0, i32 0
5096 // CHECK7-NEXT:    [[TMP8:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i8* [[TMP5]])
5097 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
5098 // CHECK7-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
5099 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A]], align 4
5100 // CHECK7-NEXT:    store i32 [[TMP9]], i32* [[A_CASTED]], align 4
5101 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[A_CASTED]], align 4
5102 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101(i32 [[TMP10]], i64* [[K]]) #[[ATTR4:[0-9]+]]
5103 // CHECK7-NEXT:    store i32 12, i32* [[LIN]], align 4
5104 // CHECK7-NEXT:    [[TMP11:%.*]] = load i16, i16* [[AA]], align 2
5105 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5106 // CHECK7-NEXT:    store i16 [[TMP11]], i16* [[CONV]], align 2
5107 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5108 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[LIN]], align 4
5109 // CHECK7-NEXT:    store i32 [[TMP13]], i32* [[LIN_CASTED]], align 4
5110 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
5111 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[A]], align 4
5112 // CHECK7-NEXT:    store i32 [[TMP15]], i32* [[A_CASTED2]], align 4
5113 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[A_CASTED2]], align 4
5114 // CHECK7-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5115 // CHECK7-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
5116 // CHECK7-NEXT:    store i32 [[TMP12]], i32* [[TMP18]], align 4
5117 // CHECK7-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5118 // CHECK7-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
5119 // CHECK7-NEXT:    store i32 [[TMP12]], i32* [[TMP20]], align 4
5120 // CHECK7-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5121 // CHECK7-NEXT:    store i8* null, i8** [[TMP21]], align 4
5122 // CHECK7-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
5123 // CHECK7-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
5124 // CHECK7-NEXT:    store i32 [[TMP14]], i32* [[TMP23]], align 4
5125 // CHECK7-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
5126 // CHECK7-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
5127 // CHECK7-NEXT:    store i32 [[TMP14]], i32* [[TMP25]], align 4
5128 // CHECK7-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
5129 // CHECK7-NEXT:    store i8* null, i8** [[TMP26]], align 4
5130 // CHECK7-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
5131 // CHECK7-NEXT:    [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
5132 // CHECK7-NEXT:    store i32 [[TMP16]], i32* [[TMP28]], align 4
5133 // CHECK7-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
5134 // CHECK7-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
5135 // CHECK7-NEXT:    store i32 [[TMP16]], i32* [[TMP30]], align 4
5136 // CHECK7-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
5137 // CHECK7-NEXT:    store i8* null, i8** [[TMP31]], align 4
5138 // CHECK7-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5139 // CHECK7-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5140 // CHECK7-NEXT:    [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108.region_id, i32 3, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
5141 // CHECK7-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
5142 // CHECK7-NEXT:    br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5143 // CHECK7:       omp_offload.failed:
5144 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108(i32 [[TMP12]], i32 [[TMP14]], i32 [[TMP16]]) #[[ATTR4]]
5145 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5146 // CHECK7:       omp_offload.cont:
5147 // CHECK7-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A]], align 4
5148 // CHECK7-NEXT:    store i32 [[TMP36]], i32* [[A_CASTED3]], align 4
5149 // CHECK7-NEXT:    [[TMP37:%.*]] = load i32, i32* [[A_CASTED3]], align 4
5150 // CHECK7-NEXT:    [[TMP38:%.*]] = load i16, i16* [[AA]], align 2
5151 // CHECK7-NEXT:    [[CONV5:%.*]] = bitcast i32* [[AA_CASTED4]] to i16*
5152 // CHECK7-NEXT:    store i16 [[TMP38]], i16* [[CONV5]], align 2
5153 // CHECK7-NEXT:    [[TMP39:%.*]] = load i32, i32* [[AA_CASTED4]], align 4
5154 // CHECK7-NEXT:    [[TMP40:%.*]] = load i32, i32* [[N_ADDR]], align 4
5155 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP40]], 10
5156 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5157 // CHECK7:       omp_if.then:
5158 // CHECK7-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
5159 // CHECK7-NEXT:    [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
5160 // CHECK7-NEXT:    store i32 [[TMP37]], i32* [[TMP42]], align 4
5161 // CHECK7-NEXT:    [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
5162 // CHECK7-NEXT:    [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
5163 // CHECK7-NEXT:    store i32 [[TMP37]], i32* [[TMP44]], align 4
5164 // CHECK7-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 0
5165 // CHECK7-NEXT:    store i8* null, i8** [[TMP45]], align 4
5166 // CHECK7-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 1
5167 // CHECK7-NEXT:    [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
5168 // CHECK7-NEXT:    store i32 [[TMP39]], i32* [[TMP47]], align 4
5169 // CHECK7-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 1
5170 // CHECK7-NEXT:    [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
5171 // CHECK7-NEXT:    store i32 [[TMP39]], i32* [[TMP49]], align 4
5172 // CHECK7-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS8]], i32 0, i32 1
5173 // CHECK7-NEXT:    store i8* null, i8** [[TMP50]], align 4
5174 // CHECK7-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS6]], i32 0, i32 0
5175 // CHECK7-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS7]], i32 0, i32 0
5176 // CHECK7-NEXT:    [[TMP53:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116.region_id, i32 2, i8** [[TMP51]], i8** [[TMP52]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.5, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.6, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
5177 // CHECK7-NEXT:    [[TMP54:%.*]] = icmp ne i32 [[TMP53]], 0
5178 // CHECK7-NEXT:    br i1 [[TMP54]], label [[OMP_OFFLOAD_FAILED9:%.*]], label [[OMP_OFFLOAD_CONT10:%.*]]
5179 // CHECK7:       omp_offload.failed9:
5180 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i32 [[TMP37]], i32 [[TMP39]]) #[[ATTR4]]
5181 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT10]]
5182 // CHECK7:       omp_offload.cont10:
5183 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
5184 // CHECK7:       omp_if.else:
5185 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116(i32 [[TMP37]], i32 [[TMP39]]) #[[ATTR4]]
5186 // CHECK7-NEXT:    br label [[OMP_IF_END]]
5187 // CHECK7:       omp_if.end:
5188 // CHECK7-NEXT:    [[TMP55:%.*]] = load i32, i32* [[A]], align 4
5189 // CHECK7-NEXT:    store i32 [[TMP55]], i32* [[DOTCAPTURE_EXPR_]], align 4
5190 // CHECK7-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A]], align 4
5191 // CHECK7-NEXT:    store i32 [[TMP56]], i32* [[A_CASTED11]], align 4
5192 // CHECK7-NEXT:    [[TMP57:%.*]] = load i32, i32* [[A_CASTED11]], align 4
5193 // CHECK7-NEXT:    [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5194 // CHECK7-NEXT:    store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5195 // CHECK7-NEXT:    [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5196 // CHECK7-NEXT:    [[TMP60:%.*]] = load i32, i32* [[N_ADDR]], align 4
5197 // CHECK7-NEXT:    [[CMP12:%.*]] = icmp sgt i32 [[TMP60]], 20
5198 // CHECK7-NEXT:    br i1 [[CMP12]], label [[OMP_IF_THEN13:%.*]], label [[OMP_IF_ELSE19:%.*]]
5199 // CHECK7:       omp_if.then13:
5200 // CHECK7-NEXT:    [[TMP61:%.*]] = mul nuw i32 [[TMP1]], 4
5201 // CHECK7-NEXT:    [[TMP62:%.*]] = sext i32 [[TMP61]] to i64
5202 // CHECK7-NEXT:    [[TMP63:%.*]] = mul nuw i32 5, [[TMP3]]
5203 // CHECK7-NEXT:    [[TMP64:%.*]] = mul nuw i32 [[TMP63]], 8
5204 // CHECK7-NEXT:    [[TMP65:%.*]] = sext i32 [[TMP64]] to i64
5205 // CHECK7-NEXT:    [[TMP66:%.*]] = bitcast [10 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
5206 // CHECK7-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP66]], i8* align 4 bitcast ([10 x i64]* @.offload_sizes.8 to i8*), i32 80, i1 false)
5207 // CHECK7-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
5208 // CHECK7-NEXT:    [[TMP68:%.*]] = bitcast i8** [[TMP67]] to i32*
5209 // CHECK7-NEXT:    store i32 [[TMP57]], i32* [[TMP68]], align 4
5210 // CHECK7-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
5211 // CHECK7-NEXT:    [[TMP70:%.*]] = bitcast i8** [[TMP69]] to i32*
5212 // CHECK7-NEXT:    store i32 [[TMP57]], i32* [[TMP70]], align 4
5213 // CHECK7-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 0
5214 // CHECK7-NEXT:    store i8* null, i8** [[TMP71]], align 4
5215 // CHECK7-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 1
5216 // CHECK7-NEXT:    [[TMP73:%.*]] = bitcast i8** [[TMP72]] to [10 x float]**
5217 // CHECK7-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP73]], align 4
5218 // CHECK7-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 1
5219 // CHECK7-NEXT:    [[TMP75:%.*]] = bitcast i8** [[TMP74]] to [10 x float]**
5220 // CHECK7-NEXT:    store [10 x float]* [[B]], [10 x float]** [[TMP75]], align 4
5221 // CHECK7-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 1
5222 // CHECK7-NEXT:    store i8* null, i8** [[TMP76]], align 4
5223 // CHECK7-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 2
5224 // CHECK7-NEXT:    [[TMP78:%.*]] = bitcast i8** [[TMP77]] to i32*
5225 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP78]], align 4
5226 // CHECK7-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 2
5227 // CHECK7-NEXT:    [[TMP80:%.*]] = bitcast i8** [[TMP79]] to i32*
5228 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP80]], align 4
5229 // CHECK7-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 2
5230 // CHECK7-NEXT:    store i8* null, i8** [[TMP81]], align 4
5231 // CHECK7-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 3
5232 // CHECK7-NEXT:    [[TMP83:%.*]] = bitcast i8** [[TMP82]] to float**
5233 // CHECK7-NEXT:    store float* [[VLA]], float** [[TMP83]], align 4
5234 // CHECK7-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 3
5235 // CHECK7-NEXT:    [[TMP85:%.*]] = bitcast i8** [[TMP84]] to float**
5236 // CHECK7-NEXT:    store float* [[VLA]], float** [[TMP85]], align 4
5237 // CHECK7-NEXT:    [[TMP86:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
5238 // CHECK7-NEXT:    store i64 [[TMP62]], i64* [[TMP86]], align 4
5239 // CHECK7-NEXT:    [[TMP87:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 3
5240 // CHECK7-NEXT:    store i8* null, i8** [[TMP87]], align 4
5241 // CHECK7-NEXT:    [[TMP88:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 4
5242 // CHECK7-NEXT:    [[TMP89:%.*]] = bitcast i8** [[TMP88]] to [5 x [10 x double]]**
5243 // CHECK7-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP89]], align 4
5244 // CHECK7-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 4
5245 // CHECK7-NEXT:    [[TMP91:%.*]] = bitcast i8** [[TMP90]] to [5 x [10 x double]]**
5246 // CHECK7-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[TMP91]], align 4
5247 // CHECK7-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 4
5248 // CHECK7-NEXT:    store i8* null, i8** [[TMP92]], align 4
5249 // CHECK7-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 5
5250 // CHECK7-NEXT:    [[TMP94:%.*]] = bitcast i8** [[TMP93]] to i32*
5251 // CHECK7-NEXT:    store i32 5, i32* [[TMP94]], align 4
5252 // CHECK7-NEXT:    [[TMP95:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 5
5253 // CHECK7-NEXT:    [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i32*
5254 // CHECK7-NEXT:    store i32 5, i32* [[TMP96]], align 4
5255 // CHECK7-NEXT:    [[TMP97:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 5
5256 // CHECK7-NEXT:    store i8* null, i8** [[TMP97]], align 4
5257 // CHECK7-NEXT:    [[TMP98:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 6
5258 // CHECK7-NEXT:    [[TMP99:%.*]] = bitcast i8** [[TMP98]] to i32*
5259 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP99]], align 4
5260 // CHECK7-NEXT:    [[TMP100:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 6
5261 // CHECK7-NEXT:    [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32*
5262 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP101]], align 4
5263 // CHECK7-NEXT:    [[TMP102:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 6
5264 // CHECK7-NEXT:    store i8* null, i8** [[TMP102]], align 4
5265 // CHECK7-NEXT:    [[TMP103:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 7
5266 // CHECK7-NEXT:    [[TMP104:%.*]] = bitcast i8** [[TMP103]] to double**
5267 // CHECK7-NEXT:    store double* [[VLA1]], double** [[TMP104]], align 4
5268 // CHECK7-NEXT:    [[TMP105:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 7
5269 // CHECK7-NEXT:    [[TMP106:%.*]] = bitcast i8** [[TMP105]] to double**
5270 // CHECK7-NEXT:    store double* [[VLA1]], double** [[TMP106]], align 4
5271 // CHECK7-NEXT:    [[TMP107:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 7
5272 // CHECK7-NEXT:    store i64 [[TMP65]], i64* [[TMP107]], align 4
5273 // CHECK7-NEXT:    [[TMP108:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 7
5274 // CHECK7-NEXT:    store i8* null, i8** [[TMP108]], align 4
5275 // CHECK7-NEXT:    [[TMP109:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 8
5276 // CHECK7-NEXT:    [[TMP110:%.*]] = bitcast i8** [[TMP109]] to %struct.TT**
5277 // CHECK7-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP110]], align 4
5278 // CHECK7-NEXT:    [[TMP111:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 8
5279 // CHECK7-NEXT:    [[TMP112:%.*]] = bitcast i8** [[TMP111]] to %struct.TT**
5280 // CHECK7-NEXT:    store %struct.TT* [[D]], %struct.TT** [[TMP112]], align 4
5281 // CHECK7-NEXT:    [[TMP113:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 8
5282 // CHECK7-NEXT:    store i8* null, i8** [[TMP113]], align 4
5283 // CHECK7-NEXT:    [[TMP114:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 9
5284 // CHECK7-NEXT:    [[TMP115:%.*]] = bitcast i8** [[TMP114]] to i32*
5285 // CHECK7-NEXT:    store i32 [[TMP59]], i32* [[TMP115]], align 4
5286 // CHECK7-NEXT:    [[TMP116:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 9
5287 // CHECK7-NEXT:    [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32*
5288 // CHECK7-NEXT:    store i32 [[TMP59]], i32* [[TMP117]], align 4
5289 // CHECK7-NEXT:    [[TMP118:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_MAPPERS16]], i32 0, i32 9
5290 // CHECK7-NEXT:    store i8* null, i8** [[TMP118]], align 4
5291 // CHECK7-NEXT:    [[TMP119:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_BASEPTRS14]], i32 0, i32 0
5292 // CHECK7-NEXT:    [[TMP120:%.*]] = getelementptr inbounds [10 x i8*], [10 x i8*]* [[DOTOFFLOAD_PTRS15]], i32 0, i32 0
5293 // CHECK7-NEXT:    [[TMP121:%.*]] = getelementptr inbounds [10 x i64], [10 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
5294 // CHECK7-NEXT:    [[TMP122:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140.region_id, i32 10, i8** [[TMP119]], i8** [[TMP120]], i64* [[TMP121]], i64* getelementptr inbounds ([10 x i64], [10 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
5295 // CHECK7-NEXT:    [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
5296 // CHECK7-NEXT:    br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
5297 // CHECK7:       omp_offload.failed17:
5298 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i32 [[TMP57]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP59]]) #[[ATTR4]]
5299 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT18]]
5300 // CHECK7:       omp_offload.cont18:
5301 // CHECK7-NEXT:    br label [[OMP_IF_END20:%.*]]
5302 // CHECK7:       omp_if.else19:
5303 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140(i32 [[TMP57]], [10 x float]* [[B]], i32 [[TMP1]], float* [[VLA]], [5 x [10 x double]]* [[C]], i32 5, i32 [[TMP3]], double* [[VLA1]], %struct.TT* [[D]], i32 [[TMP59]]) #[[ATTR4]]
5304 // CHECK7-NEXT:    br label [[OMP_IF_END20]]
5305 // CHECK7:       omp_if.end20:
5306 // CHECK7-NEXT:    [[TMP124:%.*]] = load i32, i32* [[A]], align 4
5307 // CHECK7-NEXT:    [[TMP125:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5308 // CHECK7-NEXT:    call void @llvm.stackrestore(i8* [[TMP125]])
5309 // CHECK7-NEXT:    ret i32 [[TMP124]]
5310 //
5311 //
5312 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
5313 // CHECK7-SAME: () #[[ATTR2:[0-9]+]] {
5314 // CHECK7-NEXT:  entry:
5315 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
5316 // CHECK7-NEXT:    ret void
5317 //
5318 //
5319 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
5320 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR3:[0-9]+]] {
5321 // CHECK7-NEXT:  entry:
5322 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5323 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5324 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5325 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5326 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5327 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5328 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5329 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5330 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
5331 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5332 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5333 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5334 // CHECK7-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
5335 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5336 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5337 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5338 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5339 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5340 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5341 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
5342 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5343 // CHECK7:       cond.true:
5344 // CHECK7-NEXT:    br label [[COND_END:%.*]]
5345 // CHECK7:       cond.false:
5346 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5347 // CHECK7-NEXT:    br label [[COND_END]]
5348 // CHECK7:       cond.end:
5349 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5350 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5351 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5352 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5353 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5354 // CHECK7:       omp.inner.for.cond:
5355 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
5356 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
5357 // CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5358 // CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5359 // CHECK7:       omp.inner.for.body:
5360 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
5361 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
5362 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
5363 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
5364 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5365 // CHECK7:       omp.body.continue:
5366 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5367 // CHECK7:       omp.inner.for.inc:
5368 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
5369 // CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
5370 // CHECK7-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
5371 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
5372 // CHECK7:       omp.inner.for.end:
5373 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5374 // CHECK7:       omp.loop.exit:
5375 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5376 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5377 // CHECK7-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
5378 // CHECK7-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5379 // CHECK7:       .omp.final.then:
5380 // CHECK7-NEXT:    store i32 33, i32* [[I]], align 4
5381 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5382 // CHECK7:       .omp.final.done:
5383 // CHECK7-NEXT:    ret void
5384 //
5385 //
5386 // CHECK7-LABEL: define {{[^@]+}}@.omp_task_entry.
5387 // CHECK7-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR5:[0-9]+]] {
5388 // CHECK7-NEXT:  entry:
5389 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
5390 // CHECK7-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 4
5391 // CHECK7-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 4
5392 // CHECK7-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 4
5393 // CHECK7-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 4
5394 // CHECK7-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 4
5395 // CHECK7-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
5396 // CHECK7-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 4
5397 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
5398 // CHECK7-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
5399 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
5400 // CHECK7-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 4
5401 // CHECK7-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
5402 // CHECK7-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
5403 // CHECK7-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
5404 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 4
5405 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
5406 // CHECK7-NEXT:    [[TMP9:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
5407 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META17:![0-9]+]])
5408 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META20:![0-9]+]])
5409 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META22:![0-9]+]])
5410 // CHECK7-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META24:![0-9]+]])
5411 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !26
5412 // CHECK7-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 4, !noalias !26
5413 // CHECK7-NEXT:    store i8* null, i8** [[DOTPRIVATES__ADDR_I]], align 4, !noalias !26
5414 // CHECK7-NEXT:    store void (i8*, ...)* null, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 4, !noalias !26
5415 // CHECK7-NEXT:    store i8* [[TMP9]], i8** [[DOTTASK_T__ADDR_I]], align 4, !noalias !26
5416 // CHECK7-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !26
5417 // CHECK7-NEXT:    [[TMP10:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 4, !noalias !26
5418 // CHECK7-NEXT:    [[TMP11:%.*]] = call i32 @__tgt_target_teams_nowait_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 1, i32 0, i32 0, i8* null, i32 0, i8* null) #[[ATTR4]]
5419 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5420 // CHECK7-NEXT:    br i1 [[TMP12]], label [[OMP_OFFLOAD_FAILED_I:%.*]], label [[DOTOMP_OUTLINED__1_EXIT:%.*]]
5421 // CHECK7:       omp_offload.failed.i:
5422 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96() #[[ATTR4]]
5423 // CHECK7-NEXT:    br label [[DOTOMP_OUTLINED__1_EXIT]]
5424 // CHECK7:       .omp_outlined..1.exit:
5425 // CHECK7-NEXT:    ret i32 0
5426 //
5427 //
5428 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l101
5429 // CHECK7-SAME: (i32 noundef [[A:%.*]], i64* noundef nonnull align 4 dereferenceable(8) [[K:%.*]]) #[[ATTR3]] {
5430 // CHECK7-NEXT:  entry:
5431 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5432 // CHECK7-NEXT:    [[K_ADDR:%.*]] = alloca i64*, align 4
5433 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5434 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5435 // CHECK7-NEXT:    store i64* [[K]], i64** [[K_ADDR]], align 4
5436 // CHECK7-NEXT:    [[TMP0:%.*]] = load i64*, i64** [[K_ADDR]], align 4
5437 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5438 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
5439 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
5440 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i64*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP2]], i64* [[TMP0]])
5441 // CHECK7-NEXT:    ret void
5442 //
5443 //
5444 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
5445 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i64* noundef nonnull align 4 dereferenceable(8) [[K:%.*]]) #[[ATTR3]] {
5446 // CHECK7-NEXT:  entry:
5447 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5448 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5449 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5450 // CHECK7-NEXT:    [[K_ADDR:%.*]] = alloca i64*, align 4
5451 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5452 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5453 // CHECK7-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
5454 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5455 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5456 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5457 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5458 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
5459 // CHECK7-NEXT:    [[K1:%.*]] = alloca i64, align 8
5460 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5461 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5462 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5463 // CHECK7-NEXT:    store i64* [[K]], i64** [[K_ADDR]], align 4
5464 // CHECK7-NEXT:    [[TMP0:%.*]] = load i64*, i64** [[K_ADDR]], align 4
5465 // CHECK7-NEXT:    [[TMP1:%.*]] = load i64, i64* [[TMP0]], align 8
5466 // CHECK7-NEXT:    store i64 [[TMP1]], i64* [[DOTLINEAR_START]], align 8
5467 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5468 // CHECK7-NEXT:    store i32 8, i32* [[DOTOMP_UB]], align 4
5469 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5470 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5471 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5472 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5473 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
5474 // CHECK7-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 1073741859, i32 0, i32 8, i32 1, i32 1)
5475 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
5476 // CHECK7:       omp.dispatch.cond:
5477 // CHECK7-NEXT:    [[TMP4:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
5478 // CHECK7-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP4]], 0
5479 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
5480 // CHECK7:       omp.dispatch.body:
5481 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5482 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5483 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5484 // CHECK7:       omp.inner.for.cond:
5485 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5486 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
5487 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5488 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5489 // CHECK7:       omp.inner.for.body:
5490 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5491 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
5492 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL]]
5493 // CHECK7-NEXT:    store i32 [[SUB]], i32* [[I]], align 4, !llvm.access.group !27
5494 // CHECK7-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !27
5495 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5496 // CHECK7-NEXT:    [[MUL2:%.*]] = mul nsw i32 [[TMP10]], 3
5497 // CHECK7-NEXT:    [[CONV:%.*]] = sext i32 [[MUL2]] to i64
5498 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i64 [[TMP9]], [[CONV]]
5499 // CHECK7-NEXT:    store i64 [[ADD]], i64* [[K1]], align 8, !llvm.access.group !27
5500 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !27
5501 // CHECK7-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP11]], 1
5502 // CHECK7-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !27
5503 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5504 // CHECK7:       omp.body.continue:
5505 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5506 // CHECK7:       omp.inner.for.inc:
5507 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5508 // CHECK7-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
5509 // CHECK7-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5510 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
5511 // CHECK7:       omp.inner.for.end:
5512 // CHECK7-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
5513 // CHECK7:       omp.dispatch.inc:
5514 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND]]
5515 // CHECK7:       omp.dispatch.end:
5516 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5517 // CHECK7-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5518 // CHECK7-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5519 // CHECK7:       .omp.final.then:
5520 // CHECK7-NEXT:    store i32 1, i32* [[I]], align 4
5521 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5522 // CHECK7:       .omp.final.done:
5523 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5524 // CHECK7-NEXT:    [[TMP16:%.*]] = icmp ne i32 [[TMP15]], 0
5525 // CHECK7-NEXT:    br i1 [[TMP16]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
5526 // CHECK7:       .omp.linear.pu:
5527 // CHECK7-NEXT:    [[TMP17:%.*]] = load i64, i64* [[K1]], align 8
5528 // CHECK7-NEXT:    store i64 [[TMP17]], i64* [[TMP0]], align 8
5529 // CHECK7-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
5530 // CHECK7:       .omp.linear.pu.done:
5531 // CHECK7-NEXT:    ret void
5532 //
5533 //
5534 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
5535 // CHECK7-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR2]] {
5536 // CHECK7-NEXT:  entry:
5537 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5538 // CHECK7-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
5539 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5540 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5541 // CHECK7-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
5542 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5543 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5544 // CHECK7-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
5545 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5546 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5547 // CHECK7-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
5548 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5549 // CHECK7-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
5550 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5551 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
5552 // CHECK7-NEXT:    store i32 [[TMP2]], i32* [[LIN_CASTED]], align 4
5553 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
5554 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
5555 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[A_CASTED]], align 4
5556 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A_CASTED]], align 4
5557 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]])
5558 // CHECK7-NEXT:    ret void
5559 //
5560 //
5561 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
5562 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR3]] {
5563 // CHECK7-NEXT:  entry:
5564 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5565 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5566 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5567 // CHECK7-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
5568 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5569 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
5570 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i64, align 4
5571 // CHECK7-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
5572 // CHECK7-NEXT:    [[DOTLINEAR_START1:%.*]] = alloca i32, align 4
5573 // CHECK7-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
5574 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
5575 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
5576 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
5577 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5578 // CHECK7-NEXT:    [[IT:%.*]] = alloca i64, align 8
5579 // CHECK7-NEXT:    [[LIN2:%.*]] = alloca i32, align 4
5580 // CHECK7-NEXT:    [[A3:%.*]] = alloca i32, align 4
5581 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5582 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5583 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5584 // CHECK7-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
5585 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5586 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5587 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
5588 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
5589 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
5590 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START1]], align 4
5591 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
5592 // CHECK7-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
5593 // CHECK7-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
5594 // CHECK7-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
5595 // CHECK7-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
5596 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5597 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5598 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5599 // CHECK7-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
5600 // CHECK7-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
5601 // CHECK7-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
5602 // CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
5603 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5604 // CHECK7:       cond.true:
5605 // CHECK7-NEXT:    br label [[COND_END:%.*]]
5606 // CHECK7:       cond.false:
5607 // CHECK7-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
5608 // CHECK7-NEXT:    br label [[COND_END]]
5609 // CHECK7:       cond.end:
5610 // CHECK7-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5611 // CHECK7-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
5612 // CHECK7-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
5613 // CHECK7-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
5614 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5615 // CHECK7:       omp.inner.for.cond:
5616 // CHECK7-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5617 // CHECK7-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !30
5618 // CHECK7-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
5619 // CHECK7-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5620 // CHECK7:       omp.inner.for.body:
5621 // CHECK7-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5622 // CHECK7-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
5623 // CHECK7-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
5624 // CHECK7-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !30
5625 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !30
5626 // CHECK7-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
5627 // CHECK7-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5628 // CHECK7-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !30
5629 // CHECK7-NEXT:    [[MUL6:%.*]] = mul i64 [[TMP11]], [[TMP12]]
5630 // CHECK7-NEXT:    [[ADD:%.*]] = add i64 [[CONV5]], [[MUL6]]
5631 // CHECK7-NEXT:    [[CONV7:%.*]] = trunc i64 [[ADD]] to i32
5632 // CHECK7-NEXT:    store i32 [[CONV7]], i32* [[LIN2]], align 4, !llvm.access.group !30
5633 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START1]], align 4, !llvm.access.group !30
5634 // CHECK7-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP13]] to i64
5635 // CHECK7-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5636 // CHECK7-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !30
5637 // CHECK7-NEXT:    [[MUL9:%.*]] = mul i64 [[TMP14]], [[TMP15]]
5638 // CHECK7-NEXT:    [[ADD10:%.*]] = add i64 [[CONV8]], [[MUL9]]
5639 // CHECK7-NEXT:    [[CONV11:%.*]] = trunc i64 [[ADD10]] to i32
5640 // CHECK7-NEXT:    store i32 [[CONV11]], i32* [[A3]], align 4, !llvm.access.group !30
5641 // CHECK7-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !30
5642 // CHECK7-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP16]] to i32
5643 // CHECK7-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
5644 // CHECK7-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
5645 // CHECK7-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2, !llvm.access.group !30
5646 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5647 // CHECK7:       omp.body.continue:
5648 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5649 // CHECK7:       omp.inner.for.inc:
5650 // CHECK7-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5651 // CHECK7-NEXT:    [[ADD15:%.*]] = add i64 [[TMP17]], 1
5652 // CHECK7-NEXT:    store i64 [[ADD15]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
5653 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
5654 // CHECK7:       omp.inner.for.end:
5655 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5656 // CHECK7:       omp.loop.exit:
5657 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5658 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5659 // CHECK7-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
5660 // CHECK7-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5661 // CHECK7:       .omp.final.then:
5662 // CHECK7-NEXT:    store i64 400, i64* [[IT]], align 8
5663 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5664 // CHECK7:       .omp.final.done:
5665 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5666 // CHECK7-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
5667 // CHECK7-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
5668 // CHECK7:       .omp.linear.pu:
5669 // CHECK7-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN2]], align 4
5670 // CHECK7-NEXT:    store i32 [[TMP22]], i32* [[LIN_ADDR]], align 4
5671 // CHECK7-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A3]], align 4
5672 // CHECK7-NEXT:    store i32 [[TMP23]], i32* [[A_ADDR]], align 4
5673 // CHECK7-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
5674 // CHECK7:       .omp.linear.pu.done:
5675 // CHECK7-NEXT:    ret void
5676 //
5677 //
5678 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
5679 // CHECK7-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR2]] {
5680 // CHECK7-NEXT:  entry:
5681 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5682 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5683 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5684 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
5685 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5686 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5687 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5688 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
5689 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
5690 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
5691 // CHECK7-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
5692 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
5693 // CHECK7-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
5694 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
5695 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
5696 // CHECK7-NEXT:    ret void
5697 //
5698 //
5699 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..4
5700 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR3]] {
5701 // CHECK7-NEXT:  entry:
5702 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5703 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5704 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5705 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
5706 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5707 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i16, align 2
5708 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5709 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5710 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5711 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5712 // CHECK7-NEXT:    [[IT:%.*]] = alloca i16, align 2
5713 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5714 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5715 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5716 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
5717 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
5718 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5719 // CHECK7-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
5720 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5721 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5722 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5723 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5724 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5725 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5726 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
5727 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5728 // CHECK7:       cond.true:
5729 // CHECK7-NEXT:    br label [[COND_END:%.*]]
5730 // CHECK7:       cond.false:
5731 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5732 // CHECK7-NEXT:    br label [[COND_END]]
5733 // CHECK7:       cond.end:
5734 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5735 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5736 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5737 // CHECK7-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5738 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5739 // CHECK7:       omp.inner.for.cond:
5740 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5741 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
5742 // CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5743 // CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5744 // CHECK7:       omp.inner.for.body:
5745 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5746 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
5747 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
5748 // CHECK7-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
5749 // CHECK7-NEXT:    store i16 [[CONV2]], i16* [[IT]], align 2, !llvm.access.group !33
5750 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !33
5751 // CHECK7-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
5752 // CHECK7-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !33
5753 // CHECK7-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !33
5754 // CHECK7-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
5755 // CHECK7-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
5756 // CHECK7-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
5757 // CHECK7-NEXT:    store i16 [[CONV6]], i16* [[CONV]], align 2, !llvm.access.group !33
5758 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5759 // CHECK7:       omp.body.continue:
5760 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5761 // CHECK7:       omp.inner.for.inc:
5762 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5763 // CHECK7-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
5764 // CHECK7-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5765 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
5766 // CHECK7:       omp.inner.for.end:
5767 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5768 // CHECK7:       omp.loop.exit:
5769 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5770 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5771 // CHECK7-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5772 // CHECK7-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5773 // CHECK7:       .omp.final.then:
5774 // CHECK7-NEXT:    store i16 22, i16* [[IT]], align 2
5775 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5776 // CHECK7:       .omp.final.done:
5777 // CHECK7-NEXT:    ret void
5778 //
5779 //
5780 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
5781 // CHECK7-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
5782 // CHECK7-NEXT:  entry:
5783 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5784 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
5785 // CHECK7-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5786 // CHECK7-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
5787 // CHECK7-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
5788 // CHECK7-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5789 // CHECK7-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
5790 // CHECK7-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
5791 // CHECK7-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
5792 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5793 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
5794 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
5795 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5796 // CHECK7-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
5797 // CHECK7-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5798 // CHECK7-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
5799 // CHECK7-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
5800 // CHECK7-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5801 // CHECK7-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
5802 // CHECK7-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
5803 // CHECK7-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
5804 // CHECK7-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5805 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
5806 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5807 // CHECK7-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
5808 // CHECK7-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
5809 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5810 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
5811 // CHECK7-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
5812 // CHECK7-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
5813 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
5814 // CHECK7-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
5815 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
5816 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5817 // CHECK7-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5818 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
5819 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
5820 // CHECK7-NEXT:    ret void
5821 //
5822 //
5823 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..7
5824 // CHECK7-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
5825 // CHECK7-NEXT:  entry:
5826 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5827 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5828 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
5829 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
5830 // CHECK7-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
5831 // CHECK7-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
5832 // CHECK7-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
5833 // CHECK7-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
5834 // CHECK7-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
5835 // CHECK7-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
5836 // CHECK7-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
5837 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5838 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5839 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i8, align 1
5840 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5841 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5842 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5843 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5844 // CHECK7-NEXT:    [[IT:%.*]] = alloca i8, align 1
5845 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5846 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5847 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
5848 // CHECK7-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
5849 // CHECK7-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
5850 // CHECK7-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
5851 // CHECK7-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
5852 // CHECK7-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
5853 // CHECK7-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
5854 // CHECK7-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
5855 // CHECK7-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
5856 // CHECK7-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5857 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
5858 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
5859 // CHECK7-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
5860 // CHECK7-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
5861 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
5862 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
5863 // CHECK7-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
5864 // CHECK7-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
5865 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5866 // CHECK7-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
5867 // CHECK7-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5868 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5869 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5870 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5871 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
5872 // CHECK7-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
5873 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
5874 // CHECK7:       omp.dispatch.cond:
5875 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5876 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
5877 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5878 // CHECK7:       cond.true:
5879 // CHECK7-NEXT:    br label [[COND_END:%.*]]
5880 // CHECK7:       cond.false:
5881 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5882 // CHECK7-NEXT:    br label [[COND_END]]
5883 // CHECK7:       cond.end:
5884 // CHECK7-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
5885 // CHECK7-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5886 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5887 // CHECK7-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
5888 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5889 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5890 // CHECK7-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5891 // CHECK7-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
5892 // CHECK7:       omp.dispatch.body:
5893 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5894 // CHECK7:       omp.inner.for.cond:
5895 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
5896 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !36
5897 // CHECK7-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
5898 // CHECK7-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5899 // CHECK7:       omp.inner.for.body:
5900 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
5901 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
5902 // CHECK7-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
5903 // CHECK7-NEXT:    [[CONV:%.*]] = trunc i32 [[SUB]] to i8
5904 // CHECK7-NEXT:    store i8 [[CONV]], i8* [[IT]], align 1, !llvm.access.group !36
5905 // CHECK7-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !36
5906 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
5907 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4, !llvm.access.group !36
5908 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
5909 // CHECK7-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !36
5910 // CHECK7-NEXT:    [[CONV7:%.*]] = fpext float [[TMP20]] to double
5911 // CHECK7-NEXT:    [[ADD8:%.*]] = fadd double [[CONV7]], 1.000000e+00
5912 // CHECK7-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
5913 // CHECK7-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !36
5914 // CHECK7-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
5915 // CHECK7-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !36
5916 // CHECK7-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
5917 // CHECK7-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
5918 // CHECK7-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
5919 // CHECK7-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !36
5920 // CHECK7-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
5921 // CHECK7-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
5922 // CHECK7-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !36
5923 // CHECK7-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
5924 // CHECK7-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !36
5925 // CHECK7-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
5926 // CHECK7-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
5927 // CHECK7-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
5928 // CHECK7-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !36
5929 // CHECK7-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
5930 // CHECK7-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !36
5931 // CHECK7-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
5932 // CHECK7-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !36
5933 // CHECK7-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
5934 // CHECK7-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !36
5935 // CHECK7-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
5936 // CHECK7-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !36
5937 // CHECK7-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
5938 // CHECK7-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
5939 // CHECK7-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
5940 // CHECK7-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !36
5941 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5942 // CHECK7:       omp.body.continue:
5943 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5944 // CHECK7:       omp.inner.for.inc:
5945 // CHECK7-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
5946 // CHECK7-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
5947 // CHECK7-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
5948 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
5949 // CHECK7:       omp.inner.for.end:
5950 // CHECK7-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
5951 // CHECK7:       omp.dispatch.inc:
5952 // CHECK7-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5953 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5954 // CHECK7-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
5955 // CHECK7-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
5956 // CHECK7-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5957 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5958 // CHECK7-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
5959 // CHECK7-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
5960 // CHECK7-NEXT:    br label [[OMP_DISPATCH_COND]]
5961 // CHECK7:       omp.dispatch.end:
5962 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
5963 // CHECK7-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5964 // CHECK7-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
5965 // CHECK7-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5966 // CHECK7:       .omp.final.then:
5967 // CHECK7-NEXT:    store i8 96, i8* [[IT]], align 1
5968 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5969 // CHECK7:       .omp.final.done:
5970 // CHECK7-NEXT:    ret void
5971 //
5972 //
5973 // CHECK7-LABEL: define {{[^@]+}}@_Z3bari
5974 // CHECK7-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
5975 // CHECK7-NEXT:  entry:
5976 // CHECK7-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5977 // CHECK7-NEXT:    [[A:%.*]] = alloca i32, align 4
5978 // CHECK7-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
5979 // CHECK7-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5980 // CHECK7-NEXT:    store i32 0, i32* [[A]], align 4
5981 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
5982 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
5983 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
5984 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
5985 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
5986 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
5987 // CHECK7-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
5988 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
5989 // CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
5990 // CHECK7-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
5991 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
5992 // CHECK7-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
5993 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
5994 // CHECK7-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
5995 // CHECK7-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
5996 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
5997 // CHECK7-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
5998 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
5999 // CHECK7-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
6000 // CHECK7-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
6001 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
6002 // CHECK7-NEXT:    ret i32 [[TMP8]]
6003 //
6004 //
6005 // CHECK7-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
6006 // CHECK7-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
6007 // CHECK7-NEXT:  entry:
6008 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6009 // CHECK7-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6010 // CHECK7-NEXT:    [[B:%.*]] = alloca i32, align 4
6011 // CHECK7-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
6012 // CHECK7-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
6013 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
6014 // CHECK7-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
6015 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
6016 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [6 x i8*], align 4
6017 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [6 x i8*], align 4
6018 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [6 x i8*], align 4
6019 // CHECK7-NEXT:    [[DOTOFFLOAD_SIZES:%.*]] = alloca [6 x i64], align 4
6020 // CHECK7-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6021 // CHECK7-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6022 // CHECK7-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6023 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6024 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
6025 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
6026 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
6027 // CHECK7-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
6028 // CHECK7-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
6029 // CHECK7-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
6030 // CHECK7-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
6031 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
6032 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6033 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 60
6034 // CHECK7-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
6035 // CHECK7-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
6036 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B]], align 4
6037 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[B_CASTED]], align 4
6038 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
6039 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6040 // CHECK7-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
6041 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
6042 // CHECK7-NEXT:    [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL]] to i8
6043 // CHECK7-NEXT:    store i8 [[FROMBOOL2]], i8* [[CONV]], align 1
6044 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
6045 // CHECK7-NEXT:    [[TMP9:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6046 // CHECK7-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP9]] to i1
6047 // CHECK7-NEXT:    br i1 [[TOBOOL3]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6048 // CHECK7:       omp_if.then:
6049 // CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
6050 // CHECK7-NEXT:    [[TMP10:%.*]] = mul nuw i32 2, [[TMP1]]
6051 // CHECK7-NEXT:    [[TMP11:%.*]] = mul nuw i32 [[TMP10]], 2
6052 // CHECK7-NEXT:    [[TMP12:%.*]] = sext i32 [[TMP11]] to i64
6053 // CHECK7-NEXT:    [[TMP13:%.*]] = bitcast [6 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
6054 // CHECK7-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP13]], i8* align 4 bitcast ([6 x i64]* @.offload_sizes.11 to i8*), i32 48, i1 false)
6055 // CHECK7-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6056 // CHECK7-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.S1**
6057 // CHECK7-NEXT:    store %struct.S1* [[THIS1]], %struct.S1** [[TMP15]], align 4
6058 // CHECK7-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6059 // CHECK7-NEXT:    [[TMP17:%.*]] = bitcast i8** [[TMP16]] to double**
6060 // CHECK7-NEXT:    store double* [[A]], double** [[TMP17]], align 4
6061 // CHECK7-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6062 // CHECK7-NEXT:    store i8* null, i8** [[TMP18]], align 4
6063 // CHECK7-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6064 // CHECK7-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
6065 // CHECK7-NEXT:    store i32 [[TMP6]], i32* [[TMP20]], align 4
6066 // CHECK7-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6067 // CHECK7-NEXT:    [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
6068 // CHECK7-NEXT:    store i32 [[TMP6]], i32* [[TMP22]], align 4
6069 // CHECK7-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6070 // CHECK7-NEXT:    store i8* null, i8** [[TMP23]], align 4
6071 // CHECK7-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6072 // CHECK7-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
6073 // CHECK7-NEXT:    store i32 2, i32* [[TMP25]], align 4
6074 // CHECK7-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6075 // CHECK7-NEXT:    [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
6076 // CHECK7-NEXT:    store i32 2, i32* [[TMP27]], align 4
6077 // CHECK7-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6078 // CHECK7-NEXT:    store i8* null, i8** [[TMP28]], align 4
6079 // CHECK7-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
6080 // CHECK7-NEXT:    [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
6081 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP30]], align 4
6082 // CHECK7-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
6083 // CHECK7-NEXT:    [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
6084 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP32]], align 4
6085 // CHECK7-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
6086 // CHECK7-NEXT:    store i8* null, i8** [[TMP33]], align 4
6087 // CHECK7-NEXT:    [[TMP34:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
6088 // CHECK7-NEXT:    [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i16**
6089 // CHECK7-NEXT:    store i16* [[VLA]], i16** [[TMP35]], align 4
6090 // CHECK7-NEXT:    [[TMP36:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
6091 // CHECK7-NEXT:    [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i16**
6092 // CHECK7-NEXT:    store i16* [[VLA]], i16** [[TMP37]], align 4
6093 // CHECK7-NEXT:    [[TMP38:%.*]] = getelementptr inbounds [6 x i64], [6 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
6094 // CHECK7-NEXT:    store i64 [[TMP12]], i64* [[TMP38]], align 4
6095 // CHECK7-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
6096 // CHECK7-NEXT:    store i8* null, i8** [[TMP39]], align 4
6097 // CHECK7-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 5
6098 // CHECK7-NEXT:    [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
6099 // CHECK7-NEXT:    store i32 [[TMP8]], i32* [[TMP41]], align 4
6100 // CHECK7-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 5
6101 // CHECK7-NEXT:    [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i32*
6102 // CHECK7-NEXT:    store i32 [[TMP8]], i32* [[TMP43]], align 4
6103 // CHECK7-NEXT:    [[TMP44:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 5
6104 // CHECK7-NEXT:    store i8* null, i8** [[TMP44]], align 4
6105 // CHECK7-NEXT:    [[TMP45:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6106 // CHECK7-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [6 x i8*], [6 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6107 // CHECK7-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [6 x i64], [6 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6108 // CHECK7-NEXT:    [[TMP48:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6109 // CHECK7-NEXT:    [[TOBOOL4:%.*]] = trunc i8 [[TMP48]] to i1
6110 // CHECK7-NEXT:    [[TMP49:%.*]] = select i1 [[TOBOOL4]], i32 0, i32 1
6111 // CHECK7-NEXT:    [[TMP50:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214.region_id, i32 6, i8** [[TMP45]], i8** [[TMP46]], i64* [[TMP47]], i64* getelementptr inbounds ([6 x i64], [6 x i64]* @.offload_maptypes.12, i32 0, i32 0), i8** null, i8** null, i32 1, i32 [[TMP49]])
6112 // CHECK7-NEXT:    [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
6113 // CHECK7-NEXT:    br i1 [[TMP51]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6114 // CHECK7:       omp_offload.failed:
6115 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214(%struct.S1* [[THIS1]], i32 [[TMP6]], i32 2, i32 [[TMP1]], i16* [[VLA]], i32 [[TMP8]]) #[[ATTR4]]
6116 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6117 // CHECK7:       omp_offload.cont:
6118 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6119 // CHECK7:       omp_if.else:
6120 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214(%struct.S1* [[THIS1]], i32 [[TMP6]], i32 2, i32 [[TMP1]], i16* [[VLA]], i32 [[TMP8]]) #[[ATTR4]]
6121 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6122 // CHECK7:       omp_if.end:
6123 // CHECK7-NEXT:    [[TMP52:%.*]] = mul nsw i32 1, [[TMP1]]
6124 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP52]]
6125 // CHECK7-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
6126 // CHECK7-NEXT:    [[TMP53:%.*]] = load i16, i16* [[ARRAYIDX5]], align 2
6127 // CHECK7-NEXT:    [[CONV6:%.*]] = sext i16 [[TMP53]] to i32
6128 // CHECK7-NEXT:    [[TMP54:%.*]] = load i32, i32* [[B]], align 4
6129 // CHECK7-NEXT:    [[ADD7:%.*]] = add nsw i32 [[CONV6]], [[TMP54]]
6130 // CHECK7-NEXT:    [[TMP55:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
6131 // CHECK7-NEXT:    call void @llvm.stackrestore(i8* [[TMP55]])
6132 // CHECK7-NEXT:    ret i32 [[ADD7]]
6133 //
6134 //
6135 // CHECK7-LABEL: define {{[^@]+}}@_ZL7fstatici
6136 // CHECK7-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
6137 // CHECK7-NEXT:  entry:
6138 // CHECK7-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6139 // CHECK7-NEXT:    [[A:%.*]] = alloca i32, align 4
6140 // CHECK7-NEXT:    [[AA:%.*]] = alloca i16, align 2
6141 // CHECK7-NEXT:    [[AAA:%.*]] = alloca i8, align 1
6142 // CHECK7-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
6143 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6144 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6145 // CHECK7-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
6146 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
6147 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
6148 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
6149 // CHECK7-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6150 // CHECK7-NEXT:    store i32 0, i32* [[A]], align 4
6151 // CHECK7-NEXT:    store i16 0, i16* [[AA]], align 2
6152 // CHECK7-NEXT:    store i8 0, i8* [[AAA]], align 1
6153 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
6154 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
6155 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
6156 // CHECK7-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
6157 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6158 // CHECK7-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
6159 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6160 // CHECK7-NEXT:    [[TMP4:%.*]] = load i8, i8* [[AAA]], align 1
6161 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
6162 // CHECK7-NEXT:    store i8 [[TMP4]], i8* [[CONV1]], align 1
6163 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
6164 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
6165 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 50
6166 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6167 // CHECK7:       omp_if.then:
6168 // CHECK7-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6169 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
6170 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
6171 // CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6172 // CHECK7-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
6173 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP10]], align 4
6174 // CHECK7-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6175 // CHECK7-NEXT:    store i8* null, i8** [[TMP11]], align 4
6176 // CHECK7-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6177 // CHECK7-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
6178 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
6179 // CHECK7-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6180 // CHECK7-NEXT:    [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
6181 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP15]], align 4
6182 // CHECK7-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6183 // CHECK7-NEXT:    store i8* null, i8** [[TMP16]], align 4
6184 // CHECK7-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6185 // CHECK7-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
6186 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[TMP18]], align 4
6187 // CHECK7-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6188 // CHECK7-NEXT:    [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
6189 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[TMP20]], align 4
6190 // CHECK7-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6191 // CHECK7-NEXT:    store i8* null, i8** [[TMP21]], align 4
6192 // CHECK7-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
6193 // CHECK7-NEXT:    [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [10 x i32]**
6194 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP23]], align 4
6195 // CHECK7-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
6196 // CHECK7-NEXT:    [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [10 x i32]**
6197 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP25]], align 4
6198 // CHECK7-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
6199 // CHECK7-NEXT:    store i8* null, i8** [[TMP26]], align 4
6200 // CHECK7-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6201 // CHECK7-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6202 // CHECK7-NEXT:    [[TMP29:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes.14, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.15, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6203 // CHECK7-NEXT:    [[TMP30:%.*]] = icmp ne i32 [[TMP29]], 0
6204 // CHECK7-NEXT:    br i1 [[TMP30]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6205 // CHECK7:       omp_offload.failed:
6206 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
6207 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6208 // CHECK7:       omp_offload.cont:
6209 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6210 // CHECK7:       omp_if.else:
6211 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195(i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]], [10 x i32]* [[B]]) #[[ATTR4]]
6212 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6213 // CHECK7:       omp_if.end:
6214 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[A]], align 4
6215 // CHECK7-NEXT:    ret i32 [[TMP31]]
6216 //
6217 //
6218 // CHECK7-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
6219 // CHECK7-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
6220 // CHECK7-NEXT:  entry:
6221 // CHECK7-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6222 // CHECK7-NEXT:    [[A:%.*]] = alloca i32, align 4
6223 // CHECK7-NEXT:    [[AA:%.*]] = alloca i16, align 2
6224 // CHECK7-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
6225 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6226 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6227 // CHECK7-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
6228 // CHECK7-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
6229 // CHECK7-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
6230 // CHECK7-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6231 // CHECK7-NEXT:    store i32 0, i32* [[A]], align 4
6232 // CHECK7-NEXT:    store i16 0, i16* [[AA]], align 2
6233 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
6234 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
6235 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
6236 // CHECK7-NEXT:    [[TMP2:%.*]] = load i16, i16* [[AA]], align 2
6237 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6238 // CHECK7-NEXT:    store i16 [[TMP2]], i16* [[CONV]], align 2
6239 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6240 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6241 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 40
6242 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6243 // CHECK7:       omp_if.then:
6244 // CHECK7-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6245 // CHECK7-NEXT:    [[TMP6:%.*]] = bitcast i8** [[TMP5]] to i32*
6246 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP6]], align 4
6247 // CHECK7-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6248 // CHECK7-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
6249 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[TMP8]], align 4
6250 // CHECK7-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6251 // CHECK7-NEXT:    store i8* null, i8** [[TMP9]], align 4
6252 // CHECK7-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6253 // CHECK7-NEXT:    [[TMP11:%.*]] = bitcast i8** [[TMP10]] to i32*
6254 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP11]], align 4
6255 // CHECK7-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6256 // CHECK7-NEXT:    [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
6257 // CHECK7-NEXT:    store i32 [[TMP3]], i32* [[TMP13]], align 4
6258 // CHECK7-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6259 // CHECK7-NEXT:    store i8* null, i8** [[TMP14]], align 4
6260 // CHECK7-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6261 // CHECK7-NEXT:    [[TMP16:%.*]] = bitcast i8** [[TMP15]] to [10 x i32]**
6262 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP16]], align 4
6263 // CHECK7-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6264 // CHECK7-NEXT:    [[TMP18:%.*]] = bitcast i8** [[TMP17]] to [10 x i32]**
6265 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[TMP18]], align 4
6266 // CHECK7-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6267 // CHECK7-NEXT:    store i8* null, i8** [[TMP19]], align 4
6268 // CHECK7-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6269 // CHECK7-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6270 // CHECK7-NEXT:    [[TMP22:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178.region_id, i32 3, i8** [[TMP20]], i8** [[TMP21]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 1, i32 0)
6271 // CHECK7-NEXT:    [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
6272 // CHECK7-NEXT:    br i1 [[TMP23]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6273 // CHECK7:       omp_offload.failed:
6274 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
6275 // CHECK7-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6276 // CHECK7:       omp_offload.cont:
6277 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6278 // CHECK7:       omp_if.else:
6279 // CHECK7-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[B]]) #[[ATTR4]]
6280 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6281 // CHECK7:       omp_if.end:
6282 // CHECK7-NEXT:    [[TMP24:%.*]] = load i32, i32* [[A]], align 4
6283 // CHECK7-NEXT:    ret i32 [[TMP24]]
6284 //
6285 //
6286 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214
6287 // CHECK7-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
6288 // CHECK7-NEXT:  entry:
6289 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6290 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
6291 // CHECK7-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6292 // CHECK7-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6293 // CHECK7-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
6294 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
6295 // CHECK7-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
6296 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
6297 // CHECK7-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
6298 // CHECK7-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6299 // CHECK7-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
6300 // CHECK7-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6301 // CHECK7-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
6302 // CHECK7-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6303 // CHECK7-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6304 // CHECK7-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
6305 // CHECK7-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6306 // CHECK7-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6307 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6308 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6309 // CHECK7-NEXT:    [[TMP4:%.*]] = load i16*, i16** [[C_ADDR]], align 4
6310 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
6311 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
6312 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[B_CASTED]], align 4
6313 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
6314 // CHECK7-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV]], align 1
6315 // CHECK7-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
6316 // CHECK7-NEXT:    [[CONV3:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
6317 // CHECK7-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
6318 // CHECK7-NEXT:    store i8 [[FROMBOOL]], i8* [[CONV3]], align 1
6319 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
6320 // CHECK7-NEXT:    [[TMP9:%.*]] = load i8, i8* [[CONV]], align 1
6321 // CHECK7-NEXT:    [[TOBOOL4:%.*]] = trunc i8 [[TMP9]] to i1
6322 // CHECK7-NEXT:    br i1 [[TOBOOL4]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6323 // CHECK7:       omp_if.then:
6324 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*, i32)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP6]], i32 [[TMP2]], i32 [[TMP3]], i16* [[TMP4]], i32 [[TMP8]])
6325 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6326 // CHECK7:       omp_if.else:
6327 // CHECK7-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
6328 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
6329 // CHECK7-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
6330 // CHECK7-NEXT:    call void @.omp_outlined..10(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i32 [[TMP6]], i32 [[TMP2]], i32 [[TMP3]], i16* [[TMP4]], i32 [[TMP8]]) #[[ATTR4]]
6331 // CHECK7-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
6332 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6333 // CHECK7:       omp_if.end:
6334 // CHECK7-NEXT:    ret void
6335 //
6336 //
6337 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..10
6338 // CHECK7-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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR3]] {
6339 // CHECK7-NEXT:  entry:
6340 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6341 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6342 // CHECK7-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
6343 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
6344 // CHECK7-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
6345 // CHECK7-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
6346 // CHECK7-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
6347 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
6348 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6349 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i64, align 4
6350 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6351 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6352 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6353 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6354 // CHECK7-NEXT:    [[IT:%.*]] = alloca i64, align 8
6355 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6356 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6357 // CHECK7-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
6358 // CHECK7-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
6359 // CHECK7-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6360 // CHECK7-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
6361 // CHECK7-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
6362 // CHECK7-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
6363 // CHECK7-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
6364 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6365 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
6366 // CHECK7-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
6367 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
6368 // CHECK7-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6369 // CHECK7-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
6370 // CHECK7-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6371 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6372 // CHECK7-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV]], align 1
6373 // CHECK7-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
6374 // CHECK7-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6375 // CHECK7:       omp_if.then:
6376 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6377 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
6378 // CHECK7-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
6379 // CHECK7-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6380 // CHECK7-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP7]], 3
6381 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6382 // CHECK7:       cond.true:
6383 // CHECK7-NEXT:    br label [[COND_END:%.*]]
6384 // CHECK7:       cond.false:
6385 // CHECK7-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6386 // CHECK7-NEXT:    br label [[COND_END]]
6387 // CHECK7:       cond.end:
6388 // CHECK7-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
6389 // CHECK7-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
6390 // CHECK7-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6391 // CHECK7-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_IV]], align 8
6392 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6393 // CHECK7:       omp.inner.for.cond:
6394 // CHECK7-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
6395 // CHECK7-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !39
6396 // CHECK7-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP10]], [[TMP11]]
6397 // CHECK7-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6398 // CHECK7:       omp.inner.for.body:
6399 // CHECK7-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
6400 // CHECK7-NEXT:    [[MUL:%.*]] = mul i64 [[TMP12]], 400
6401 // CHECK7-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
6402 // CHECK7-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !39
6403 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[B_ADDR]], align 4, !llvm.access.group !39
6404 // CHECK7-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP13]] to double
6405 // CHECK7-NEXT:    [[ADD:%.*]] = fadd double [[CONV4]], 1.500000e+00
6406 // CHECK7-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
6407 // CHECK7-NEXT:    store double [[ADD]], double* [[A]], align 4, !nontemporal !40, !llvm.access.group !39
6408 // CHECK7-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6409 // CHECK7-NEXT:    [[TMP14:%.*]] = load double, double* [[A5]], align 4, !nontemporal !40, !llvm.access.group !39
6410 // CHECK7-NEXT:    [[INC:%.*]] = fadd double [[TMP14]], 1.000000e+00
6411 // CHECK7-NEXT:    store double [[INC]], double* [[A5]], align 4, !nontemporal !40, !llvm.access.group !39
6412 // CHECK7-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
6413 // CHECK7-NEXT:    [[TMP15:%.*]] = mul nsw i32 1, [[TMP2]]
6414 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP15]]
6415 // CHECK7-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
6416 // CHECK7-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2, !llvm.access.group !39
6417 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6418 // CHECK7:       omp.body.continue:
6419 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6420 // CHECK7:       omp.inner.for.inc:
6421 // CHECK7-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
6422 // CHECK7-NEXT:    [[ADD8:%.*]] = add i64 [[TMP16]], 1
6423 // CHECK7-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !39
6424 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
6425 // CHECK7:       omp.inner.for.end:
6426 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6427 // CHECK7:       omp_if.else:
6428 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6429 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
6430 // CHECK7-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
6431 // CHECK7-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6432 // CHECK7-NEXT:    [[CMP9:%.*]] = icmp ugt i64 [[TMP19]], 3
6433 // CHECK7-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
6434 // CHECK7:       cond.true10:
6435 // CHECK7-NEXT:    br label [[COND_END12:%.*]]
6436 // CHECK7:       cond.false11:
6437 // CHECK7-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6438 // CHECK7-NEXT:    br label [[COND_END12]]
6439 // CHECK7:       cond.end12:
6440 // CHECK7-NEXT:    [[COND13:%.*]] = phi i64 [ 3, [[COND_TRUE10]] ], [ [[TMP20]], [[COND_FALSE11]] ]
6441 // CHECK7-NEXT:    store i64 [[COND13]], i64* [[DOTOMP_UB]], align 8
6442 // CHECK7-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6443 // CHECK7-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV]], align 8
6444 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND14:%.*]]
6445 // CHECK7:       omp.inner.for.cond14:
6446 // CHECK7-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6447 // CHECK7-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6448 // CHECK7-NEXT:    [[CMP15:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
6449 // CHECK7-NEXT:    br i1 [[CMP15]], label [[OMP_INNER_FOR_BODY16:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
6450 // CHECK7:       omp.inner.for.body16:
6451 // CHECK7-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6452 // CHECK7-NEXT:    [[MUL17:%.*]] = mul i64 [[TMP24]], 400
6453 // CHECK7-NEXT:    [[SUB18:%.*]] = sub i64 2000, [[MUL17]]
6454 // CHECK7-NEXT:    store i64 [[SUB18]], i64* [[IT]], align 8
6455 // CHECK7-NEXT:    [[TMP25:%.*]] = load i32, i32* [[B_ADDR]], align 4
6456 // CHECK7-NEXT:    [[CONV19:%.*]] = sitofp i32 [[TMP25]] to double
6457 // CHECK7-NEXT:    [[ADD20:%.*]] = fadd double [[CONV19]], 1.500000e+00
6458 // CHECK7-NEXT:    [[A21:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6459 // CHECK7-NEXT:    store double [[ADD20]], double* [[A21]], align 4
6460 // CHECK7-NEXT:    [[A22:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
6461 // CHECK7-NEXT:    [[TMP26:%.*]] = load double, double* [[A22]], align 4
6462 // CHECK7-NEXT:    [[INC23:%.*]] = fadd double [[TMP26]], 1.000000e+00
6463 // CHECK7-NEXT:    store double [[INC23]], double* [[A22]], align 4
6464 // CHECK7-NEXT:    [[CONV24:%.*]] = fptosi double [[INC23]] to i16
6465 // CHECK7-NEXT:    [[TMP27:%.*]] = mul nsw i32 1, [[TMP2]]
6466 // CHECK7-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP27]]
6467 // CHECK7-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX25]], i32 1
6468 // CHECK7-NEXT:    store i16 [[CONV24]], i16* [[ARRAYIDX26]], align 2
6469 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
6470 // CHECK7:       omp.body.continue27:
6471 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
6472 // CHECK7:       omp.inner.for.inc28:
6473 // CHECK7-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6474 // CHECK7-NEXT:    [[ADD29:%.*]] = add i64 [[TMP28]], 1
6475 // CHECK7-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8
6476 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND14]], !llvm.loop [[LOOP43:![0-9]+]]
6477 // CHECK7:       omp.inner.for.end30:
6478 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6479 // CHECK7:       omp_if.end:
6480 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6481 // CHECK7:       omp.loop.exit:
6482 // CHECK7-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6483 // CHECK7-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
6484 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
6485 // CHECK7-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6486 // CHECK7-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
6487 // CHECK7-NEXT:    br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6488 // CHECK7:       .omp.final.then:
6489 // CHECK7-NEXT:    store i64 400, i64* [[IT]], align 8
6490 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6491 // CHECK7:       .omp.final.done:
6492 // CHECK7-NEXT:    ret void
6493 //
6494 //
6495 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
6496 // CHECK7-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6497 // CHECK7-NEXT:  entry:
6498 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6499 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6500 // CHECK7-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
6501 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6502 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6503 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6504 // CHECK7-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
6505 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6506 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6507 // CHECK7-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
6508 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6509 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6510 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
6511 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6512 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
6513 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
6514 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
6515 // CHECK7-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
6516 // CHECK7-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6517 // CHECK7-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
6518 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6519 // CHECK7-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
6520 // CHECK7-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
6521 // CHECK7-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
6522 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
6523 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..13 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
6524 // CHECK7-NEXT:    ret void
6525 //
6526 //
6527 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..13
6528 // CHECK7-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:%.*]]) #[[ATTR3]] {
6529 // CHECK7-NEXT:  entry:
6530 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6531 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6532 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6533 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6534 // CHECK7-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
6535 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6536 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6537 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6538 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6539 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6540 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6541 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6542 // CHECK7-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
6543 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6544 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6545 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
6546 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6547 // CHECK7-NEXT:    ret void
6548 //
6549 //
6550 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
6551 // CHECK7-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR2]] {
6552 // CHECK7-NEXT:  entry:
6553 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6554 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6555 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6556 // CHECK7-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
6557 // CHECK7-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
6558 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6559 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6560 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6561 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6562 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6563 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
6564 // CHECK7-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
6565 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
6566 // CHECK7-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
6567 // CHECK7-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
6568 // CHECK7-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
6569 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
6570 // CHECK7-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
6571 // CHECK7-NEXT:    ret void
6572 //
6573 //
6574 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..16
6575 // CHECK7-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:%.*]]) #[[ATTR3]] {
6576 // CHECK7-NEXT:  entry:
6577 // CHECK7-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6578 // CHECK7-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6579 // CHECK7-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
6580 // CHECK7-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
6581 // CHECK7-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
6582 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6583 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i64, align 4
6584 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6585 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6586 // CHECK7-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6587 // CHECK7-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6588 // CHECK7-NEXT:    [[I:%.*]] = alloca i64, align 8
6589 // CHECK7-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6590 // CHECK7-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6591 // CHECK7-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
6592 // CHECK7-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
6593 // CHECK7-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
6594 // CHECK7-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
6595 // CHECK7-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
6596 // CHECK7-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6597 // CHECK7-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
6598 // CHECK7-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6599 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6600 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6601 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6602 // CHECK7-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
6603 // CHECK7-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6604 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
6605 // CHECK7-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6606 // CHECK7:       cond.true:
6607 // CHECK7-NEXT:    br label [[COND_END:%.*]]
6608 // CHECK7:       cond.false:
6609 // CHECK7-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6610 // CHECK7-NEXT:    br label [[COND_END]]
6611 // CHECK7:       cond.end:
6612 // CHECK7-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6613 // CHECK7-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
6614 // CHECK7-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6615 // CHECK7-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
6616 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6617 // CHECK7:       omp.inner.for.cond:
6618 // CHECK7-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !45
6619 // CHECK7-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !45
6620 // CHECK7-NEXT:    [[CMP1:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
6621 // CHECK7-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6622 // CHECK7:       omp.inner.for.body:
6623 // CHECK7-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !45
6624 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
6625 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
6626 // CHECK7-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !45
6627 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !45
6628 // CHECK7-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
6629 // CHECK7-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4, !llvm.access.group !45
6630 // CHECK7-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !45
6631 // CHECK7-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
6632 // CHECK7-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
6633 // CHECK7-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
6634 // CHECK7-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2, !llvm.access.group !45
6635 // CHECK7-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
6636 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !45
6637 // CHECK7-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
6638 // CHECK7-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !45
6639 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6640 // CHECK7:       omp.body.continue:
6641 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6642 // CHECK7:       omp.inner.for.inc:
6643 // CHECK7-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !45
6644 // CHECK7-NEXT:    [[ADD7:%.*]] = add nsw i64 [[TMP12]], 1
6645 // CHECK7-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !45
6646 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
6647 // CHECK7:       omp.inner.for.end:
6648 // CHECK7-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6649 // CHECK7:       omp.loop.exit:
6650 // CHECK7-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6651 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6652 // CHECK7-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6653 // CHECK7-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6654 // CHECK7:       .omp.final.then:
6655 // CHECK7-NEXT:    store i64 11, i64* [[I]], align 8
6656 // CHECK7-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6657 // CHECK7:       .omp.final.done:
6658 // CHECK7-NEXT:    ret void
6659 //
6660 //
6661 // CHECK7-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6662 // CHECK7-SAME: () #[[ATTR8:[0-9]+]] {
6663 // CHECK7-NEXT:  entry:
6664 // CHECK7-NEXT:    call void @__tgt_register_requires(i64 1)
6665 // CHECK7-NEXT:    ret void
6666 //
6667 //
6668 // CHECK9-LABEL: define {{[^@]+}}@_Z7get_valv
6669 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
6670 // CHECK9-NEXT:  entry:
6671 // CHECK9-NEXT:    ret i64 0
6672 //
6673 //
6674 // CHECK9-LABEL: define {{[^@]+}}@_Z3fooi
6675 // CHECK9-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
6676 // CHECK9-NEXT:  entry:
6677 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6678 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
6679 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
6680 // CHECK9-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
6681 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
6682 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
6683 // CHECK9-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
6684 // CHECK9-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
6685 // CHECK9-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
6686 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6687 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6688 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6689 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6690 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
6691 // CHECK9-NEXT:    [[K:%.*]] = alloca i64, align 8
6692 // CHECK9-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
6693 // CHECK9-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
6694 // CHECK9-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
6695 // CHECK9-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
6696 // CHECK9-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
6697 // CHECK9-NEXT:    [[I7:%.*]] = alloca i32, align 4
6698 // CHECK9-NEXT:    [[K8:%.*]] = alloca i64, align 8
6699 // CHECK9-NEXT:    [[LIN:%.*]] = alloca i32, align 4
6700 // CHECK9-NEXT:    [[_TMP20:%.*]] = alloca i64, align 8
6701 // CHECK9-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i64, align 8
6702 // CHECK9-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i64, align 8
6703 // CHECK9-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i64, align 8
6704 // CHECK9-NEXT:    [[DOTLINEAR_START24:%.*]] = alloca i32, align 4
6705 // CHECK9-NEXT:    [[DOTLINEAR_START25:%.*]] = alloca i32, align 4
6706 // CHECK9-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
6707 // CHECK9-NEXT:    [[IT:%.*]] = alloca i64, align 8
6708 // CHECK9-NEXT:    [[LIN27:%.*]] = alloca i32, align 4
6709 // CHECK9-NEXT:    [[A28:%.*]] = alloca i32, align 4
6710 // CHECK9-NEXT:    [[_TMP49:%.*]] = alloca i16, align 2
6711 // CHECK9-NEXT:    [[DOTOMP_LB50:%.*]] = alloca i32, align 4
6712 // CHECK9-NEXT:    [[DOTOMP_UB51:%.*]] = alloca i32, align 4
6713 // CHECK9-NEXT:    [[DOTOMP_IV52:%.*]] = alloca i32, align 4
6714 // CHECK9-NEXT:    [[IT53:%.*]] = alloca i16, align 2
6715 // CHECK9-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6716 // CHECK9-NEXT:    [[_TMP68:%.*]] = alloca i8, align 1
6717 // CHECK9-NEXT:    [[DOTOMP_LB69:%.*]] = alloca i32, align 4
6718 // CHECK9-NEXT:    [[DOTOMP_UB70:%.*]] = alloca i32, align 4
6719 // CHECK9-NEXT:    [[DOTOMP_IV71:%.*]] = alloca i32, align 4
6720 // CHECK9-NEXT:    [[IT72:%.*]] = alloca i8, align 1
6721 // CHECK9-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6722 // CHECK9-NEXT:    store i32 0, i32* [[A]], align 4
6723 // CHECK9-NEXT:    store i16 0, i16* [[AA]], align 2
6724 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6725 // CHECK9-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
6726 // CHECK9-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
6727 // CHECK9-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
6728 // CHECK9-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 4
6729 // CHECK9-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
6730 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
6731 // CHECK9-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
6732 // CHECK9-NEXT:    [[TMP5:%.*]] = mul nuw i64 5, [[TMP4]]
6733 // CHECK9-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP5]], align 8
6734 // CHECK9-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR1]], align 8
6735 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6736 // CHECK9-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
6737 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6738 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6739 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6740 // CHECK9:       omp.inner.for.cond:
6741 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6742 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
6743 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6744 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6745 // CHECK9:       omp.inner.for.body:
6746 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6747 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 5
6748 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
6749 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
6750 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6751 // CHECK9:       omp.body.continue:
6752 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6753 // CHECK9:       omp.inner.for.inc:
6754 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6755 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
6756 // CHECK9-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6757 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
6758 // CHECK9:       omp.inner.for.end:
6759 // CHECK9-NEXT:    store i32 33, i32* [[I]], align 4
6760 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
6761 // CHECK9-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
6762 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
6763 // CHECK9-NEXT:    store i32 8, i32* [[DOTOMP_UB5]], align 4
6764 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
6765 // CHECK9-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV6]], align 4
6766 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, i64* [[K]], align 8
6767 // CHECK9-NEXT:    store i64 [[TMP12]], i64* [[DOTLINEAR_START]], align 8
6768 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND9:%.*]]
6769 // CHECK9:       omp.inner.for.cond9:
6770 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
6771 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
6772 // CHECK9-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
6773 // CHECK9-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
6774 // CHECK9:       omp.inner.for.body11:
6775 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
6776 // CHECK9-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[TMP15]], 1
6777 // CHECK9-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL12]]
6778 // CHECK9-NEXT:    store i32 [[SUB]], i32* [[I7]], align 4, !llvm.access.group !6
6779 // CHECK9-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !6
6780 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
6781 // CHECK9-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP17]], 3
6782 // CHECK9-NEXT:    [[CONV:%.*]] = sext i32 [[MUL13]] to i64
6783 // CHECK9-NEXT:    [[ADD14:%.*]] = add nsw i64 [[TMP16]], [[CONV]]
6784 // CHECK9-NEXT:    store i64 [[ADD14]], i64* [[K8]], align 8, !llvm.access.group !6
6785 // CHECK9-NEXT:    [[TMP18:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !6
6786 // CHECK9-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP18]], 1
6787 // CHECK9-NEXT:    store i32 [[ADD15]], i32* [[A]], align 4, !llvm.access.group !6
6788 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
6789 // CHECK9:       omp.body.continue16:
6790 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
6791 // CHECK9:       omp.inner.for.inc17:
6792 // CHECK9-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
6793 // CHECK9-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP19]], 1
6794 // CHECK9-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
6795 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP7:![0-9]+]]
6796 // CHECK9:       omp.inner.for.end19:
6797 // CHECK9-NEXT:    store i32 1, i32* [[I7]], align 4
6798 // CHECK9-NEXT:    [[TMP20:%.*]] = load i64, i64* [[K8]], align 8
6799 // CHECK9-NEXT:    store i64 [[TMP20]], i64* [[K]], align 8
6800 // CHECK9-NEXT:    store i32 12, i32* [[LIN]], align 4
6801 // CHECK9-NEXT:    store i64 0, i64* [[DOTOMP_LB21]], align 8
6802 // CHECK9-NEXT:    store i64 3, i64* [[DOTOMP_UB22]], align 8
6803 // CHECK9-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB21]], align 8
6804 // CHECK9-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV23]], align 8
6805 // CHECK9-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN]], align 4
6806 // CHECK9-NEXT:    store i32 [[TMP22]], i32* [[DOTLINEAR_START24]], align 4
6807 // CHECK9-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A]], align 4
6808 // CHECK9-NEXT:    store i32 [[TMP23]], i32* [[DOTLINEAR_START25]], align 4
6809 // CHECK9-NEXT:    [[CALL26:%.*]] = call noundef i64 @_Z7get_valv()
6810 // CHECK9-NEXT:    store i64 [[CALL26]], i64* [[DOTLINEAR_STEP]], align 8
6811 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
6812 // CHECK9:       omp.inner.for.cond29:
6813 // CHECK9-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6814 // CHECK9-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_UB22]], align 8, !llvm.access.group !9
6815 // CHECK9-NEXT:    [[CMP30:%.*]] = icmp ule i64 [[TMP24]], [[TMP25]]
6816 // CHECK9-NEXT:    br i1 [[CMP30]], label [[OMP_INNER_FOR_BODY31:%.*]], label [[OMP_INNER_FOR_END48:%.*]]
6817 // CHECK9:       omp.inner.for.body31:
6818 // CHECK9-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6819 // CHECK9-NEXT:    [[MUL32:%.*]] = mul i64 [[TMP26]], 400
6820 // CHECK9-NEXT:    [[SUB33:%.*]] = sub i64 2000, [[MUL32]]
6821 // CHECK9-NEXT:    store i64 [[SUB33]], i64* [[IT]], align 8, !llvm.access.group !9
6822 // CHECK9-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTLINEAR_START24]], align 4, !llvm.access.group !9
6823 // CHECK9-NEXT:    [[CONV34:%.*]] = sext i32 [[TMP27]] to i64
6824 // CHECK9-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6825 // CHECK9-NEXT:    [[TMP29:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !9
6826 // CHECK9-NEXT:    [[MUL35:%.*]] = mul i64 [[TMP28]], [[TMP29]]
6827 // CHECK9-NEXT:    [[ADD36:%.*]] = add i64 [[CONV34]], [[MUL35]]
6828 // CHECK9-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
6829 // CHECK9-NEXT:    store i32 [[CONV37]], i32* [[LIN27]], align 4, !llvm.access.group !9
6830 // CHECK9-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTLINEAR_START25]], align 4, !llvm.access.group !9
6831 // CHECK9-NEXT:    [[CONV38:%.*]] = sext i32 [[TMP30]] to i64
6832 // CHECK9-NEXT:    [[TMP31:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6833 // CHECK9-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !9
6834 // CHECK9-NEXT:    [[MUL39:%.*]] = mul i64 [[TMP31]], [[TMP32]]
6835 // CHECK9-NEXT:    [[ADD40:%.*]] = add i64 [[CONV38]], [[MUL39]]
6836 // CHECK9-NEXT:    [[CONV41:%.*]] = trunc i64 [[ADD40]] to i32
6837 // CHECK9-NEXT:    store i32 [[CONV41]], i32* [[A28]], align 4, !llvm.access.group !9
6838 // CHECK9-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !9
6839 // CHECK9-NEXT:    [[CONV42:%.*]] = sext i16 [[TMP33]] to i32
6840 // CHECK9-NEXT:    [[ADD43:%.*]] = add nsw i32 [[CONV42]], 1
6841 // CHECK9-NEXT:    [[CONV44:%.*]] = trunc i32 [[ADD43]] to i16
6842 // CHECK9-NEXT:    store i16 [[CONV44]], i16* [[AA]], align 2, !llvm.access.group !9
6843 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE45:%.*]]
6844 // CHECK9:       omp.body.continue45:
6845 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC46:%.*]]
6846 // CHECK9:       omp.inner.for.inc46:
6847 // CHECK9-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6848 // CHECK9-NEXT:    [[ADD47:%.*]] = add i64 [[TMP34]], 1
6849 // CHECK9-NEXT:    store i64 [[ADD47]], i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
6850 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP10:![0-9]+]]
6851 // CHECK9:       omp.inner.for.end48:
6852 // CHECK9-NEXT:    store i64 400, i64* [[IT]], align 8
6853 // CHECK9-NEXT:    [[TMP35:%.*]] = load i32, i32* [[LIN27]], align 4
6854 // CHECK9-NEXT:    store i32 [[TMP35]], i32* [[LIN]], align 4
6855 // CHECK9-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A28]], align 4
6856 // CHECK9-NEXT:    store i32 [[TMP36]], i32* [[A]], align 4
6857 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB50]], align 4
6858 // CHECK9-NEXT:    store i32 3, i32* [[DOTOMP_UB51]], align 4
6859 // CHECK9-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_LB50]], align 4
6860 // CHECK9-NEXT:    store i32 [[TMP37]], i32* [[DOTOMP_IV52]], align 4
6861 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND54:%.*]]
6862 // CHECK9:       omp.inner.for.cond54:
6863 // CHECK9-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
6864 // CHECK9-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_UB51]], align 4, !llvm.access.group !12
6865 // CHECK9-NEXT:    [[CMP55:%.*]] = icmp sle i32 [[TMP38]], [[TMP39]]
6866 // CHECK9-NEXT:    br i1 [[CMP55]], label [[OMP_INNER_FOR_BODY56:%.*]], label [[OMP_INNER_FOR_END67:%.*]]
6867 // CHECK9:       omp.inner.for.body56:
6868 // CHECK9-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
6869 // CHECK9-NEXT:    [[MUL57:%.*]] = mul nsw i32 [[TMP40]], 4
6870 // CHECK9-NEXT:    [[ADD58:%.*]] = add nsw i32 6, [[MUL57]]
6871 // CHECK9-NEXT:    [[CONV59:%.*]] = trunc i32 [[ADD58]] to i16
6872 // CHECK9-NEXT:    store i16 [[CONV59]], i16* [[IT53]], align 2, !llvm.access.group !12
6873 // CHECK9-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !12
6874 // CHECK9-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP41]], 1
6875 // CHECK9-NEXT:    store i32 [[ADD60]], i32* [[A]], align 4, !llvm.access.group !12
6876 // CHECK9-NEXT:    [[TMP42:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !12
6877 // CHECK9-NEXT:    [[CONV61:%.*]] = sext i16 [[TMP42]] to i32
6878 // CHECK9-NEXT:    [[ADD62:%.*]] = add nsw i32 [[CONV61]], 1
6879 // CHECK9-NEXT:    [[CONV63:%.*]] = trunc i32 [[ADD62]] to i16
6880 // CHECK9-NEXT:    store i16 [[CONV63]], i16* [[AA]], align 2, !llvm.access.group !12
6881 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE64:%.*]]
6882 // CHECK9:       omp.body.continue64:
6883 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC65:%.*]]
6884 // CHECK9:       omp.inner.for.inc65:
6885 // CHECK9-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
6886 // CHECK9-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP43]], 1
6887 // CHECK9-NEXT:    store i32 [[ADD66]], i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
6888 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND54]], !llvm.loop [[LOOP13:![0-9]+]]
6889 // CHECK9:       omp.inner.for.end67:
6890 // CHECK9-NEXT:    store i16 22, i16* [[IT53]], align 2
6891 // CHECK9-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
6892 // CHECK9-NEXT:    store i32 [[TMP44]], i32* [[DOTCAPTURE_EXPR_]], align 4
6893 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB69]], align 4
6894 // CHECK9-NEXT:    store i32 25, i32* [[DOTOMP_UB70]], align 4
6895 // CHECK9-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_LB69]], align 4
6896 // CHECK9-NEXT:    store i32 [[TMP45]], i32* [[DOTOMP_IV71]], align 4
6897 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND73:%.*]]
6898 // CHECK9:       omp.inner.for.cond73:
6899 // CHECK9-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
6900 // CHECK9-NEXT:    [[TMP47:%.*]] = load i32, i32* [[DOTOMP_UB70]], align 4, !llvm.access.group !15
6901 // CHECK9-NEXT:    [[CMP74:%.*]] = icmp sle i32 [[TMP46]], [[TMP47]]
6902 // CHECK9-NEXT:    br i1 [[CMP74]], label [[OMP_INNER_FOR_BODY75:%.*]], label [[OMP_INNER_FOR_END100:%.*]]
6903 // CHECK9:       omp.inner.for.body75:
6904 // CHECK9-NEXT:    [[TMP48:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
6905 // CHECK9-NEXT:    [[MUL76:%.*]] = mul nsw i32 [[TMP48]], 1
6906 // CHECK9-NEXT:    [[SUB77:%.*]] = sub nsw i32 122, [[MUL76]]
6907 // CHECK9-NEXT:    [[CONV78:%.*]] = trunc i32 [[SUB77]] to i8
6908 // CHECK9-NEXT:    store i8 [[CONV78]], i8* [[IT72]], align 1, !llvm.access.group !15
6909 // CHECK9-NEXT:    [[TMP49:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !15
6910 // CHECK9-NEXT:    [[ADD79:%.*]] = add nsw i32 [[TMP49]], 1
6911 // CHECK9-NEXT:    store i32 [[ADD79]], i32* [[A]], align 4, !llvm.access.group !15
6912 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[B]], i64 0, i64 2
6913 // CHECK9-NEXT:    [[TMP50:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !15
6914 // CHECK9-NEXT:    [[CONV80:%.*]] = fpext float [[TMP50]] to double
6915 // CHECK9-NEXT:    [[ADD81:%.*]] = fadd double [[CONV80]], 1.000000e+00
6916 // CHECK9-NEXT:    [[CONV82:%.*]] = fptrunc double [[ADD81]] to float
6917 // CHECK9-NEXT:    store float [[CONV82]], float* [[ARRAYIDX]], align 4, !llvm.access.group !15
6918 // CHECK9-NEXT:    [[ARRAYIDX83:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 3
6919 // CHECK9-NEXT:    [[TMP51:%.*]] = load float, float* [[ARRAYIDX83]], align 4, !llvm.access.group !15
6920 // CHECK9-NEXT:    [[CONV84:%.*]] = fpext float [[TMP51]] to double
6921 // CHECK9-NEXT:    [[ADD85:%.*]] = fadd double [[CONV84]], 1.000000e+00
6922 // CHECK9-NEXT:    [[CONV86:%.*]] = fptrunc double [[ADD85]] to float
6923 // CHECK9-NEXT:    store float [[CONV86]], float* [[ARRAYIDX83]], align 4, !llvm.access.group !15
6924 // CHECK9-NEXT:    [[ARRAYIDX87:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[C]], i64 0, i64 1
6925 // CHECK9-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX87]], i64 0, i64 2
6926 // CHECK9-NEXT:    [[TMP52:%.*]] = load double, double* [[ARRAYIDX88]], align 8, !llvm.access.group !15
6927 // CHECK9-NEXT:    [[ADD89:%.*]] = fadd double [[TMP52]], 1.000000e+00
6928 // CHECK9-NEXT:    store double [[ADD89]], double* [[ARRAYIDX88]], align 8, !llvm.access.group !15
6929 // CHECK9-NEXT:    [[TMP53:%.*]] = mul nsw i64 1, [[TMP4]]
6930 // CHECK9-NEXT:    [[ARRAYIDX90:%.*]] = getelementptr inbounds double, double* [[VLA1]], i64 [[TMP53]]
6931 // CHECK9-NEXT:    [[ARRAYIDX91:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX90]], i64 3
6932 // CHECK9-NEXT:    [[TMP54:%.*]] = load double, double* [[ARRAYIDX91]], align 8, !llvm.access.group !15
6933 // CHECK9-NEXT:    [[ADD92:%.*]] = fadd double [[TMP54]], 1.000000e+00
6934 // CHECK9-NEXT:    store double [[ADD92]], double* [[ARRAYIDX91]], align 8, !llvm.access.group !15
6935 // CHECK9-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 0
6936 // CHECK9-NEXT:    [[TMP55:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !15
6937 // CHECK9-NEXT:    [[ADD93:%.*]] = add nsw i64 [[TMP55]], 1
6938 // CHECK9-NEXT:    store i64 [[ADD93]], i64* [[X]], align 8, !llvm.access.group !15
6939 // CHECK9-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 1
6940 // CHECK9-NEXT:    [[TMP56:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !15
6941 // CHECK9-NEXT:    [[CONV94:%.*]] = sext i8 [[TMP56]] to i32
6942 // CHECK9-NEXT:    [[ADD95:%.*]] = add nsw i32 [[CONV94]], 1
6943 // CHECK9-NEXT:    [[CONV96:%.*]] = trunc i32 [[ADD95]] to i8
6944 // CHECK9-NEXT:    store i8 [[CONV96]], i8* [[Y]], align 8, !llvm.access.group !15
6945 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE97:%.*]]
6946 // CHECK9:       omp.body.continue97:
6947 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC98:%.*]]
6948 // CHECK9:       omp.inner.for.inc98:
6949 // CHECK9-NEXT:    [[TMP57:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
6950 // CHECK9-NEXT:    [[ADD99:%.*]] = add nsw i32 [[TMP57]], 1
6951 // CHECK9-NEXT:    store i32 [[ADD99]], i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
6952 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND73]], !llvm.loop [[LOOP16:![0-9]+]]
6953 // CHECK9:       omp.inner.for.end100:
6954 // CHECK9-NEXT:    store i8 96, i8* [[IT72]], align 1
6955 // CHECK9-NEXT:    [[TMP58:%.*]] = load i32, i32* [[A]], align 4
6956 // CHECK9-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
6957 // CHECK9-NEXT:    call void @llvm.stackrestore(i8* [[TMP59]])
6958 // CHECK9-NEXT:    ret i32 [[TMP58]]
6959 //
6960 //
6961 // CHECK9-LABEL: define {{[^@]+}}@_Z3bari
6962 // CHECK9-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
6963 // CHECK9-NEXT:  entry:
6964 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6965 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
6966 // CHECK9-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
6967 // CHECK9-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
6968 // CHECK9-NEXT:    store i32 0, i32* [[A]], align 4
6969 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
6970 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
6971 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
6972 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
6973 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
6974 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
6975 // CHECK9-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
6976 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
6977 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
6978 // CHECK9-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
6979 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
6980 // CHECK9-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
6981 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
6982 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
6983 // CHECK9-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
6984 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
6985 // CHECK9-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
6986 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
6987 // CHECK9-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
6988 // CHECK9-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
6989 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
6990 // CHECK9-NEXT:    ret i32 [[TMP8]]
6991 //
6992 //
6993 // CHECK9-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
6994 // CHECK9-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
6995 // CHECK9-NEXT:  entry:
6996 // CHECK9-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
6997 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
6998 // CHECK9-NEXT:    [[B:%.*]] = alloca i32, align 4
6999 // CHECK9-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7000 // CHECK9-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7001 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i64, align 8
7002 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7003 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7004 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7005 // CHECK9-NEXT:    [[IT:%.*]] = alloca i64, align 8
7006 // CHECK9-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7007 // CHECK9-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7008 // CHECK9-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7009 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7010 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
7011 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
7012 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7013 // CHECK9-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
7014 // CHECK9-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
7015 // CHECK9-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
7016 // CHECK9-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
7017 // CHECK9-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
7018 // CHECK9-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
7019 // CHECK9-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7020 // CHECK9-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
7021 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7022 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
7023 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7024 // CHECK9:       omp.inner.for.cond:
7025 // CHECK9-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7026 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !18
7027 // CHECK9-NEXT:    [[CMP:%.*]] = icmp ule i64 [[TMP6]], [[TMP7]]
7028 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7029 // CHECK9:       omp.inner.for.body:
7030 // CHECK9-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7031 // CHECK9-NEXT:    [[MUL:%.*]] = mul i64 [[TMP8]], 400
7032 // CHECK9-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
7033 // CHECK9-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !18
7034 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[B]], align 4, !llvm.access.group !18
7035 // CHECK9-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP9]] to double
7036 // CHECK9-NEXT:    [[ADD2:%.*]] = fadd double [[CONV]], 1.500000e+00
7037 // CHECK9-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
7038 // CHECK9-NEXT:    store double [[ADD2]], double* [[A]], align 8, !llvm.access.group !18
7039 // CHECK9-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
7040 // CHECK9-NEXT:    [[TMP10:%.*]] = load double, double* [[A3]], align 8, !llvm.access.group !18
7041 // CHECK9-NEXT:    [[INC:%.*]] = fadd double [[TMP10]], 1.000000e+00
7042 // CHECK9-NEXT:    store double [[INC]], double* [[A3]], align 8, !llvm.access.group !18
7043 // CHECK9-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
7044 // CHECK9-NEXT:    [[TMP11:%.*]] = mul nsw i64 1, [[TMP2]]
7045 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP11]]
7046 // CHECK9-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
7047 // CHECK9-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2, !llvm.access.group !18
7048 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7049 // CHECK9:       omp.body.continue:
7050 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7051 // CHECK9:       omp.inner.for.inc:
7052 // CHECK9-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7053 // CHECK9-NEXT:    [[ADD6:%.*]] = add i64 [[TMP12]], 1
7054 // CHECK9-NEXT:    store i64 [[ADD6]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7055 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
7056 // CHECK9:       omp.inner.for.end:
7057 // CHECK9-NEXT:    store i64 400, i64* [[IT]], align 8
7058 // CHECK9-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP2]]
7059 // CHECK9-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP13]]
7060 // CHECK9-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX7]], i64 1
7061 // CHECK9-NEXT:    [[TMP14:%.*]] = load i16, i16* [[ARRAYIDX8]], align 2
7062 // CHECK9-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP14]] to i32
7063 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[B]], align 4
7064 // CHECK9-NEXT:    [[ADD10:%.*]] = add nsw i32 [[CONV9]], [[TMP15]]
7065 // CHECK9-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7066 // CHECK9-NEXT:    call void @llvm.stackrestore(i8* [[TMP16]])
7067 // CHECK9-NEXT:    ret i32 [[ADD10]]
7068 //
7069 //
7070 // CHECK9-LABEL: define {{[^@]+}}@_ZL7fstatici
7071 // CHECK9-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
7072 // CHECK9-NEXT:  entry:
7073 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7074 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
7075 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
7076 // CHECK9-NEXT:    [[AAA:%.*]] = alloca i8, align 1
7077 // CHECK9-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7078 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7079 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7080 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7081 // CHECK9-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7082 // CHECK9-NEXT:    store i32 0, i32* [[A]], align 4
7083 // CHECK9-NEXT:    store i16 0, i16* [[AA]], align 2
7084 // CHECK9-NEXT:    store i8 0, i8* [[AAA]], align 1
7085 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7086 // CHECK9-NEXT:    store i32 429496720, i32* [[DOTOMP_UB]], align 4
7087 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
7088 // CHECK9-NEXT:    ret i32 [[TMP0]]
7089 //
7090 //
7091 // CHECK9-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
7092 // CHECK9-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
7093 // CHECK9-NEXT:  entry:
7094 // CHECK9-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7095 // CHECK9-NEXT:    [[A:%.*]] = alloca i32, align 4
7096 // CHECK9-NEXT:    [[AA:%.*]] = alloca i16, align 2
7097 // CHECK9-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7098 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i64, align 8
7099 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7100 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7101 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7102 // CHECK9-NEXT:    [[I:%.*]] = alloca i64, align 8
7103 // CHECK9-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7104 // CHECK9-NEXT:    store i32 0, i32* [[A]], align 4
7105 // CHECK9-NEXT:    store i16 0, i16* [[AA]], align 2
7106 // CHECK9-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7107 // CHECK9-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
7108 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7109 // CHECK9-NEXT:    store i64 [[TMP0]], i64* [[DOTOMP_IV]], align 8
7110 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7111 // CHECK9:       omp.inner.for.cond:
7112 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !21
7113 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !21
7114 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sle i64 [[TMP1]], [[TMP2]]
7115 // CHECK9-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7116 // CHECK9:       omp.inner.for.body:
7117 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !21
7118 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP3]], 3
7119 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
7120 // CHECK9-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !21
7121 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !21
7122 // CHECK9-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7123 // CHECK9-NEXT:    store i32 [[ADD1]], i32* [[A]], align 4, !llvm.access.group !21
7124 // CHECK9-NEXT:    [[TMP5:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !21
7125 // CHECK9-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
7126 // CHECK9-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV]], 1
7127 // CHECK9-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
7128 // CHECK9-NEXT:    store i16 [[CONV3]], i16* [[AA]], align 2, !llvm.access.group !21
7129 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[B]], i64 0, i64 2
7130 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
7131 // CHECK9-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP6]], 1
7132 // CHECK9-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
7133 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7134 // CHECK9:       omp.body.continue:
7135 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7136 // CHECK9:       omp.inner.for.inc:
7137 // CHECK9-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !21
7138 // CHECK9-NEXT:    [[ADD5:%.*]] = add nsw i64 [[TMP7]], 1
7139 // CHECK9-NEXT:    store i64 [[ADD5]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !21
7140 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
7141 // CHECK9:       omp.inner.for.end:
7142 // CHECK9-NEXT:    store i64 11, i64* [[I]], align 8
7143 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7144 // CHECK9-NEXT:    ret i32 [[TMP8]]
7145 //
7146 //
7147 // CHECK11-LABEL: define {{[^@]+}}@_Z7get_valv
7148 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
7149 // CHECK11-NEXT:  entry:
7150 // CHECK11-NEXT:    ret i64 0
7151 //
7152 //
7153 // CHECK11-LABEL: define {{[^@]+}}@_Z3fooi
7154 // CHECK11-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
7155 // CHECK11-NEXT:  entry:
7156 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7157 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
7158 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
7159 // CHECK11-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
7160 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
7161 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
7162 // CHECK11-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
7163 // CHECK11-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
7164 // CHECK11-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
7165 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7166 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7167 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7168 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7169 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
7170 // CHECK11-NEXT:    [[K:%.*]] = alloca i64, align 8
7171 // CHECK11-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
7172 // CHECK11-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
7173 // CHECK11-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
7174 // CHECK11-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
7175 // CHECK11-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
7176 // CHECK11-NEXT:    [[I7:%.*]] = alloca i32, align 4
7177 // CHECK11-NEXT:    [[K8:%.*]] = alloca i64, align 8
7178 // CHECK11-NEXT:    [[LIN:%.*]] = alloca i32, align 4
7179 // CHECK11-NEXT:    [[_TMP20:%.*]] = alloca i64, align 4
7180 // CHECK11-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i64, align 8
7181 // CHECK11-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i64, align 8
7182 // CHECK11-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i64, align 8
7183 // CHECK11-NEXT:    [[DOTLINEAR_START24:%.*]] = alloca i32, align 4
7184 // CHECK11-NEXT:    [[DOTLINEAR_START25:%.*]] = alloca i32, align 4
7185 // CHECK11-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
7186 // CHECK11-NEXT:    [[IT:%.*]] = alloca i64, align 8
7187 // CHECK11-NEXT:    [[LIN27:%.*]] = alloca i32, align 4
7188 // CHECK11-NEXT:    [[A28:%.*]] = alloca i32, align 4
7189 // CHECK11-NEXT:    [[_TMP49:%.*]] = alloca i16, align 2
7190 // CHECK11-NEXT:    [[DOTOMP_LB50:%.*]] = alloca i32, align 4
7191 // CHECK11-NEXT:    [[DOTOMP_UB51:%.*]] = alloca i32, align 4
7192 // CHECK11-NEXT:    [[DOTOMP_IV52:%.*]] = alloca i32, align 4
7193 // CHECK11-NEXT:    [[IT53:%.*]] = alloca i16, align 2
7194 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7195 // CHECK11-NEXT:    [[_TMP68:%.*]] = alloca i8, align 1
7196 // CHECK11-NEXT:    [[DOTOMP_LB69:%.*]] = alloca i32, align 4
7197 // CHECK11-NEXT:    [[DOTOMP_UB70:%.*]] = alloca i32, align 4
7198 // CHECK11-NEXT:    [[DOTOMP_IV71:%.*]] = alloca i32, align 4
7199 // CHECK11-NEXT:    [[IT72:%.*]] = alloca i8, align 1
7200 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7201 // CHECK11-NEXT:    store i32 0, i32* [[A]], align 4
7202 // CHECK11-NEXT:    store i16 0, i16* [[AA]], align 2
7203 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7204 // CHECK11-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
7205 // CHECK11-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
7206 // CHECK11-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP0]], align 4
7207 // CHECK11-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
7208 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
7209 // CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 5, [[TMP2]]
7210 // CHECK11-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP3]], align 8
7211 // CHECK11-NEXT:    store i32 [[TMP2]], i32* [[__VLA_EXPR1]], align 4
7212 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7213 // CHECK11-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
7214 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7215 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7216 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7217 // CHECK11:       omp.inner.for.cond:
7218 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
7219 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
7220 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7221 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7222 // CHECK11:       omp.inner.for.body:
7223 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
7224 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
7225 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
7226 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
7227 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7228 // CHECK11:       omp.body.continue:
7229 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7230 // CHECK11:       omp.inner.for.inc:
7231 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
7232 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
7233 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
7234 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
7235 // CHECK11:       omp.inner.for.end:
7236 // CHECK11-NEXT:    store i32 33, i32* [[I]], align 4
7237 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
7238 // CHECK11-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
7239 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
7240 // CHECK11-NEXT:    store i32 8, i32* [[DOTOMP_UB5]], align 4
7241 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
7242 // CHECK11-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV6]], align 4
7243 // CHECK11-NEXT:    [[TMP10:%.*]] = load i64, i64* [[K]], align 8
7244 // CHECK11-NEXT:    store i64 [[TMP10]], i64* [[DOTLINEAR_START]], align 8
7245 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND9:%.*]]
7246 // CHECK11:       omp.inner.for.cond9:
7247 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
7248 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
7249 // CHECK11-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
7250 // CHECK11-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
7251 // CHECK11:       omp.inner.for.body11:
7252 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
7253 // CHECK11-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[TMP13]], 1
7254 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL12]]
7255 // CHECK11-NEXT:    store i32 [[SUB]], i32* [[I7]], align 4, !llvm.access.group !7
7256 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !7
7257 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
7258 // CHECK11-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP15]], 3
7259 // CHECK11-NEXT:    [[CONV:%.*]] = sext i32 [[MUL13]] to i64
7260 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i64 [[TMP14]], [[CONV]]
7261 // CHECK11-NEXT:    store i64 [[ADD14]], i64* [[K8]], align 8, !llvm.access.group !7
7262 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !7
7263 // CHECK11-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP16]], 1
7264 // CHECK11-NEXT:    store i32 [[ADD15]], i32* [[A]], align 4, !llvm.access.group !7
7265 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
7266 // CHECK11:       omp.body.continue16:
7267 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
7268 // CHECK11:       omp.inner.for.inc17:
7269 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
7270 // CHECK11-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP17]], 1
7271 // CHECK11-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
7272 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP8:![0-9]+]]
7273 // CHECK11:       omp.inner.for.end19:
7274 // CHECK11-NEXT:    store i32 1, i32* [[I7]], align 4
7275 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, i64* [[K8]], align 8
7276 // CHECK11-NEXT:    store i64 [[TMP18]], i64* [[K]], align 8
7277 // CHECK11-NEXT:    store i32 12, i32* [[LIN]], align 4
7278 // CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB21]], align 8
7279 // CHECK11-NEXT:    store i64 3, i64* [[DOTOMP_UB22]], align 8
7280 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_LB21]], align 8
7281 // CHECK11-NEXT:    store i64 [[TMP19]], i64* [[DOTOMP_IV23]], align 8
7282 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[LIN]], align 4
7283 // CHECK11-NEXT:    store i32 [[TMP20]], i32* [[DOTLINEAR_START24]], align 4
7284 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[A]], align 4
7285 // CHECK11-NEXT:    store i32 [[TMP21]], i32* [[DOTLINEAR_START25]], align 4
7286 // CHECK11-NEXT:    [[CALL26:%.*]] = call noundef i64 @_Z7get_valv()
7287 // CHECK11-NEXT:    store i64 [[CALL26]], i64* [[DOTLINEAR_STEP]], align 8
7288 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
7289 // CHECK11:       omp.inner.for.cond29:
7290 // CHECK11-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7291 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB22]], align 8, !llvm.access.group !10
7292 // CHECK11-NEXT:    [[CMP30:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
7293 // CHECK11-NEXT:    br i1 [[CMP30]], label [[OMP_INNER_FOR_BODY31:%.*]], label [[OMP_INNER_FOR_END48:%.*]]
7294 // CHECK11:       omp.inner.for.body31:
7295 // CHECK11-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7296 // CHECK11-NEXT:    [[MUL32:%.*]] = mul i64 [[TMP24]], 400
7297 // CHECK11-NEXT:    [[SUB33:%.*]] = sub i64 2000, [[MUL32]]
7298 // CHECK11-NEXT:    store i64 [[SUB33]], i64* [[IT]], align 8, !llvm.access.group !10
7299 // CHECK11-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTLINEAR_START24]], align 4, !llvm.access.group !10
7300 // CHECK11-NEXT:    [[CONV34:%.*]] = sext i32 [[TMP25]] to i64
7301 // CHECK11-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7302 // CHECK11-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !10
7303 // CHECK11-NEXT:    [[MUL35:%.*]] = mul i64 [[TMP26]], [[TMP27]]
7304 // CHECK11-NEXT:    [[ADD36:%.*]] = add i64 [[CONV34]], [[MUL35]]
7305 // CHECK11-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
7306 // CHECK11-NEXT:    store i32 [[CONV37]], i32* [[LIN27]], align 4, !llvm.access.group !10
7307 // CHECK11-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTLINEAR_START25]], align 4, !llvm.access.group !10
7308 // CHECK11-NEXT:    [[CONV38:%.*]] = sext i32 [[TMP28]] to i64
7309 // CHECK11-NEXT:    [[TMP29:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7310 // CHECK11-NEXT:    [[TMP30:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !10
7311 // CHECK11-NEXT:    [[MUL39:%.*]] = mul i64 [[TMP29]], [[TMP30]]
7312 // CHECK11-NEXT:    [[ADD40:%.*]] = add i64 [[CONV38]], [[MUL39]]
7313 // CHECK11-NEXT:    [[CONV41:%.*]] = trunc i64 [[ADD40]] to i32
7314 // CHECK11-NEXT:    store i32 [[CONV41]], i32* [[A28]], align 4, !llvm.access.group !10
7315 // CHECK11-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !10
7316 // CHECK11-NEXT:    [[CONV42:%.*]] = sext i16 [[TMP31]] to i32
7317 // CHECK11-NEXT:    [[ADD43:%.*]] = add nsw i32 [[CONV42]], 1
7318 // CHECK11-NEXT:    [[CONV44:%.*]] = trunc i32 [[ADD43]] to i16
7319 // CHECK11-NEXT:    store i16 [[CONV44]], i16* [[AA]], align 2, !llvm.access.group !10
7320 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE45:%.*]]
7321 // CHECK11:       omp.body.continue45:
7322 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC46:%.*]]
7323 // CHECK11:       omp.inner.for.inc46:
7324 // CHECK11-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7325 // CHECK11-NEXT:    [[ADD47:%.*]] = add i64 [[TMP32]], 1
7326 // CHECK11-NEXT:    store i64 [[ADD47]], i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
7327 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP11:![0-9]+]]
7328 // CHECK11:       omp.inner.for.end48:
7329 // CHECK11-NEXT:    store i64 400, i64* [[IT]], align 8
7330 // CHECK11-NEXT:    [[TMP33:%.*]] = load i32, i32* [[LIN27]], align 4
7331 // CHECK11-NEXT:    store i32 [[TMP33]], i32* [[LIN]], align 4
7332 // CHECK11-NEXT:    [[TMP34:%.*]] = load i32, i32* [[A28]], align 4
7333 // CHECK11-NEXT:    store i32 [[TMP34]], i32* [[A]], align 4
7334 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB50]], align 4
7335 // CHECK11-NEXT:    store i32 3, i32* [[DOTOMP_UB51]], align 4
7336 // CHECK11-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_LB50]], align 4
7337 // CHECK11-NEXT:    store i32 [[TMP35]], i32* [[DOTOMP_IV52]], align 4
7338 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND54:%.*]]
7339 // CHECK11:       omp.inner.for.cond54:
7340 // CHECK11-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
7341 // CHECK11-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_UB51]], align 4, !llvm.access.group !13
7342 // CHECK11-NEXT:    [[CMP55:%.*]] = icmp sle i32 [[TMP36]], [[TMP37]]
7343 // CHECK11-NEXT:    br i1 [[CMP55]], label [[OMP_INNER_FOR_BODY56:%.*]], label [[OMP_INNER_FOR_END67:%.*]]
7344 // CHECK11:       omp.inner.for.body56:
7345 // CHECK11-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
7346 // CHECK11-NEXT:    [[MUL57:%.*]] = mul nsw i32 [[TMP38]], 4
7347 // CHECK11-NEXT:    [[ADD58:%.*]] = add nsw i32 6, [[MUL57]]
7348 // CHECK11-NEXT:    [[CONV59:%.*]] = trunc i32 [[ADD58]] to i16
7349 // CHECK11-NEXT:    store i16 [[CONV59]], i16* [[IT53]], align 2, !llvm.access.group !13
7350 // CHECK11-NEXT:    [[TMP39:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !13
7351 // CHECK11-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP39]], 1
7352 // CHECK11-NEXT:    store i32 [[ADD60]], i32* [[A]], align 4, !llvm.access.group !13
7353 // CHECK11-NEXT:    [[TMP40:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !13
7354 // CHECK11-NEXT:    [[CONV61:%.*]] = sext i16 [[TMP40]] to i32
7355 // CHECK11-NEXT:    [[ADD62:%.*]] = add nsw i32 [[CONV61]], 1
7356 // CHECK11-NEXT:    [[CONV63:%.*]] = trunc i32 [[ADD62]] to i16
7357 // CHECK11-NEXT:    store i16 [[CONV63]], i16* [[AA]], align 2, !llvm.access.group !13
7358 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE64:%.*]]
7359 // CHECK11:       omp.body.continue64:
7360 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC65:%.*]]
7361 // CHECK11:       omp.inner.for.inc65:
7362 // CHECK11-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
7363 // CHECK11-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP41]], 1
7364 // CHECK11-NEXT:    store i32 [[ADD66]], i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
7365 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND54]], !llvm.loop [[LOOP14:![0-9]+]]
7366 // CHECK11:       omp.inner.for.end67:
7367 // CHECK11-NEXT:    store i16 22, i16* [[IT53]], align 2
7368 // CHECK11-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A]], align 4
7369 // CHECK11-NEXT:    store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_]], align 4
7370 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB69]], align 4
7371 // CHECK11-NEXT:    store i32 25, i32* [[DOTOMP_UB70]], align 4
7372 // CHECK11-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_LB69]], align 4
7373 // CHECK11-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV71]], align 4
7374 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND73:%.*]]
7375 // CHECK11:       omp.inner.for.cond73:
7376 // CHECK11-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
7377 // CHECK11-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_UB70]], align 4, !llvm.access.group !16
7378 // CHECK11-NEXT:    [[CMP74:%.*]] = icmp sle i32 [[TMP44]], [[TMP45]]
7379 // CHECK11-NEXT:    br i1 [[CMP74]], label [[OMP_INNER_FOR_BODY75:%.*]], label [[OMP_INNER_FOR_END100:%.*]]
7380 // CHECK11:       omp.inner.for.body75:
7381 // CHECK11-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
7382 // CHECK11-NEXT:    [[MUL76:%.*]] = mul nsw i32 [[TMP46]], 1
7383 // CHECK11-NEXT:    [[SUB77:%.*]] = sub nsw i32 122, [[MUL76]]
7384 // CHECK11-NEXT:    [[CONV78:%.*]] = trunc i32 [[SUB77]] to i8
7385 // CHECK11-NEXT:    store i8 [[CONV78]], i8* [[IT72]], align 1, !llvm.access.group !16
7386 // CHECK11-NEXT:    [[TMP47:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !16
7387 // CHECK11-NEXT:    [[ADD79:%.*]] = add nsw i32 [[TMP47]], 1
7388 // CHECK11-NEXT:    store i32 [[ADD79]], i32* [[A]], align 4, !llvm.access.group !16
7389 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[B]], i32 0, i32 2
7390 // CHECK11-NEXT:    [[TMP48:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !16
7391 // CHECK11-NEXT:    [[CONV80:%.*]] = fpext float [[TMP48]] to double
7392 // CHECK11-NEXT:    [[ADD81:%.*]] = fadd double [[CONV80]], 1.000000e+00
7393 // CHECK11-NEXT:    [[CONV82:%.*]] = fptrunc double [[ADD81]] to float
7394 // CHECK11-NEXT:    store float [[CONV82]], float* [[ARRAYIDX]], align 4, !llvm.access.group !16
7395 // CHECK11-NEXT:    [[ARRAYIDX83:%.*]] = getelementptr inbounds float, float* [[VLA]], i32 3
7396 // CHECK11-NEXT:    [[TMP49:%.*]] = load float, float* [[ARRAYIDX83]], align 4, !llvm.access.group !16
7397 // CHECK11-NEXT:    [[CONV84:%.*]] = fpext float [[TMP49]] to double
7398 // CHECK11-NEXT:    [[ADD85:%.*]] = fadd double [[CONV84]], 1.000000e+00
7399 // CHECK11-NEXT:    [[CONV86:%.*]] = fptrunc double [[ADD85]] to float
7400 // CHECK11-NEXT:    store float [[CONV86]], float* [[ARRAYIDX83]], align 4, !llvm.access.group !16
7401 // CHECK11-NEXT:    [[ARRAYIDX87:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[C]], i32 0, i32 1
7402 // CHECK11-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX87]], i32 0, i32 2
7403 // CHECK11-NEXT:    [[TMP50:%.*]] = load double, double* [[ARRAYIDX88]], align 8, !llvm.access.group !16
7404 // CHECK11-NEXT:    [[ADD89:%.*]] = fadd double [[TMP50]], 1.000000e+00
7405 // CHECK11-NEXT:    store double [[ADD89]], double* [[ARRAYIDX88]], align 8, !llvm.access.group !16
7406 // CHECK11-NEXT:    [[TMP51:%.*]] = mul nsw i32 1, [[TMP2]]
7407 // CHECK11-NEXT:    [[ARRAYIDX90:%.*]] = getelementptr inbounds double, double* [[VLA1]], i32 [[TMP51]]
7408 // CHECK11-NEXT:    [[ARRAYIDX91:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX90]], i32 3
7409 // CHECK11-NEXT:    [[TMP52:%.*]] = load double, double* [[ARRAYIDX91]], align 8, !llvm.access.group !16
7410 // CHECK11-NEXT:    [[ADD92:%.*]] = fadd double [[TMP52]], 1.000000e+00
7411 // CHECK11-NEXT:    store double [[ADD92]], double* [[ARRAYIDX91]], align 8, !llvm.access.group !16
7412 // CHECK11-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 0
7413 // CHECK11-NEXT:    [[TMP53:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !16
7414 // CHECK11-NEXT:    [[ADD93:%.*]] = add nsw i64 [[TMP53]], 1
7415 // CHECK11-NEXT:    store i64 [[ADD93]], i64* [[X]], align 4, !llvm.access.group !16
7416 // CHECK11-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 1
7417 // CHECK11-NEXT:    [[TMP54:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !16
7418 // CHECK11-NEXT:    [[CONV94:%.*]] = sext i8 [[TMP54]] to i32
7419 // CHECK11-NEXT:    [[ADD95:%.*]] = add nsw i32 [[CONV94]], 1
7420 // CHECK11-NEXT:    [[CONV96:%.*]] = trunc i32 [[ADD95]] to i8
7421 // CHECK11-NEXT:    store i8 [[CONV96]], i8* [[Y]], align 4, !llvm.access.group !16
7422 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE97:%.*]]
7423 // CHECK11:       omp.body.continue97:
7424 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC98:%.*]]
7425 // CHECK11:       omp.inner.for.inc98:
7426 // CHECK11-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
7427 // CHECK11-NEXT:    [[ADD99:%.*]] = add nsw i32 [[TMP55]], 1
7428 // CHECK11-NEXT:    store i32 [[ADD99]], i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
7429 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND73]], !llvm.loop [[LOOP17:![0-9]+]]
7430 // CHECK11:       omp.inner.for.end100:
7431 // CHECK11-NEXT:    store i8 96, i8* [[IT72]], align 1
7432 // CHECK11-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A]], align 4
7433 // CHECK11-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
7434 // CHECK11-NEXT:    call void @llvm.stackrestore(i8* [[TMP57]])
7435 // CHECK11-NEXT:    ret i32 [[TMP56]]
7436 //
7437 //
7438 // CHECK11-LABEL: define {{[^@]+}}@_Z3bari
7439 // CHECK11-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
7440 // CHECK11-NEXT:  entry:
7441 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7442 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
7443 // CHECK11-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
7444 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7445 // CHECK11-NEXT:    store i32 0, i32* [[A]], align 4
7446 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7447 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
7448 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
7449 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
7450 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7451 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
7452 // CHECK11-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
7453 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
7454 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
7455 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
7456 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7457 // CHECK11-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
7458 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
7459 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
7460 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
7461 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
7462 // CHECK11-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
7463 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
7464 // CHECK11-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
7465 // CHECK11-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
7466 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7467 // CHECK11-NEXT:    ret i32 [[TMP8]]
7468 //
7469 //
7470 // CHECK11-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
7471 // CHECK11-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
7472 // CHECK11-NEXT:  entry:
7473 // CHECK11-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
7474 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7475 // CHECK11-NEXT:    [[B:%.*]] = alloca i32, align 4
7476 // CHECK11-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
7477 // CHECK11-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
7478 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i64, align 4
7479 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7480 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7481 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7482 // CHECK11-NEXT:    [[IT:%.*]] = alloca i64, align 8
7483 // CHECK11-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
7484 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7485 // CHECK11-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
7486 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7487 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
7488 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
7489 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7490 // CHECK11-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
7491 // CHECK11-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
7492 // CHECK11-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
7493 // CHECK11-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
7494 // CHECK11-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
7495 // CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7496 // CHECK11-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
7497 // CHECK11-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7498 // CHECK11-NEXT:    store i64 [[TMP4]], i64* [[DOTOMP_IV]], align 8
7499 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7500 // CHECK11:       omp.inner.for.cond:
7501 // CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
7502 // CHECK11-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !19
7503 // CHECK11-NEXT:    [[CMP:%.*]] = icmp ule i64 [[TMP5]], [[TMP6]]
7504 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7505 // CHECK11:       omp.inner.for.body:
7506 // CHECK11-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
7507 // CHECK11-NEXT:    [[MUL:%.*]] = mul i64 [[TMP7]], 400
7508 // CHECK11-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
7509 // CHECK11-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !19
7510 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[B]], align 4, !llvm.access.group !19
7511 // CHECK11-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP8]] to double
7512 // CHECK11-NEXT:    [[ADD2:%.*]] = fadd double [[CONV]], 1.500000e+00
7513 // CHECK11-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
7514 // CHECK11-NEXT:    store double [[ADD2]], double* [[A]], align 4, !llvm.access.group !19
7515 // CHECK11-NEXT:    [[A3:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
7516 // CHECK11-NEXT:    [[TMP9:%.*]] = load double, double* [[A3]], align 4, !llvm.access.group !19
7517 // CHECK11-NEXT:    [[INC:%.*]] = fadd double [[TMP9]], 1.000000e+00
7518 // CHECK11-NEXT:    store double [[INC]], double* [[A3]], align 4, !llvm.access.group !19
7519 // CHECK11-NEXT:    [[CONV4:%.*]] = fptosi double [[INC]] to i16
7520 // CHECK11-NEXT:    [[TMP10:%.*]] = mul nsw i32 1, [[TMP1]]
7521 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP10]]
7522 // CHECK11-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
7523 // CHECK11-NEXT:    store i16 [[CONV4]], i16* [[ARRAYIDX5]], align 2, !llvm.access.group !19
7524 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7525 // CHECK11:       omp.body.continue:
7526 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7527 // CHECK11:       omp.inner.for.inc:
7528 // CHECK11-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
7529 // CHECK11-NEXT:    [[ADD6:%.*]] = add i64 [[TMP11]], 1
7530 // CHECK11-NEXT:    store i64 [[ADD6]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
7531 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
7532 // CHECK11:       omp.inner.for.end:
7533 // CHECK11-NEXT:    store i64 400, i64* [[IT]], align 8
7534 // CHECK11-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP1]]
7535 // CHECK11-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP12]]
7536 // CHECK11-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX7]], i32 1
7537 // CHECK11-NEXT:    [[TMP13:%.*]] = load i16, i16* [[ARRAYIDX8]], align 2
7538 // CHECK11-NEXT:    [[CONV9:%.*]] = sext i16 [[TMP13]] to i32
7539 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[B]], align 4
7540 // CHECK11-NEXT:    [[ADD10:%.*]] = add nsw i32 [[CONV9]], [[TMP14]]
7541 // CHECK11-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
7542 // CHECK11-NEXT:    call void @llvm.stackrestore(i8* [[TMP15]])
7543 // CHECK11-NEXT:    ret i32 [[ADD10]]
7544 //
7545 //
7546 // CHECK11-LABEL: define {{[^@]+}}@_ZL7fstatici
7547 // CHECK11-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
7548 // CHECK11-NEXT:  entry:
7549 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7550 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
7551 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
7552 // CHECK11-NEXT:    [[AAA:%.*]] = alloca i8, align 1
7553 // CHECK11-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7554 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7555 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7556 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7557 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7558 // CHECK11-NEXT:    store i32 0, i32* [[A]], align 4
7559 // CHECK11-NEXT:    store i16 0, i16* [[AA]], align 2
7560 // CHECK11-NEXT:    store i8 0, i8* [[AAA]], align 1
7561 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7562 // CHECK11-NEXT:    store i32 429496720, i32* [[DOTOMP_UB]], align 4
7563 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
7564 // CHECK11-NEXT:    ret i32 [[TMP0]]
7565 //
7566 //
7567 // CHECK11-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
7568 // CHECK11-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
7569 // CHECK11-NEXT:  entry:
7570 // CHECK11-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7571 // CHECK11-NEXT:    [[A:%.*]] = alloca i32, align 4
7572 // CHECK11-NEXT:    [[AA:%.*]] = alloca i16, align 2
7573 // CHECK11-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
7574 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i64, align 4
7575 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7576 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7577 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7578 // CHECK11-NEXT:    [[I:%.*]] = alloca i64, align 8
7579 // CHECK11-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7580 // CHECK11-NEXT:    store i32 0, i32* [[A]], align 4
7581 // CHECK11-NEXT:    store i16 0, i16* [[AA]], align 2
7582 // CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7583 // CHECK11-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
7584 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7585 // CHECK11-NEXT:    store i64 [[TMP0]], i64* [[DOTOMP_IV]], align 8
7586 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7587 // CHECK11:       omp.inner.for.cond:
7588 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !22
7589 // CHECK11-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !22
7590 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sle i64 [[TMP1]], [[TMP2]]
7591 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7592 // CHECK11:       omp.inner.for.body:
7593 // CHECK11-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !22
7594 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP3]], 3
7595 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
7596 // CHECK11-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !22
7597 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !22
7598 // CHECK11-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7599 // CHECK11-NEXT:    store i32 [[ADD1]], i32* [[A]], align 4, !llvm.access.group !22
7600 // CHECK11-NEXT:    [[TMP5:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !22
7601 // CHECK11-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
7602 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV]], 1
7603 // CHECK11-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
7604 // CHECK11-NEXT:    store i16 [[CONV3]], i16* [[AA]], align 2, !llvm.access.group !22
7605 // CHECK11-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[B]], i32 0, i32 2
7606 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
7607 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP6]], 1
7608 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
7609 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7610 // CHECK11:       omp.body.continue:
7611 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7612 // CHECK11:       omp.inner.for.inc:
7613 // CHECK11-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !22
7614 // CHECK11-NEXT:    [[ADD5:%.*]] = add nsw i64 [[TMP7]], 1
7615 // CHECK11-NEXT:    store i64 [[ADD5]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !22
7616 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
7617 // CHECK11:       omp.inner.for.end:
7618 // CHECK11-NEXT:    store i64 11, i64* [[I]], align 8
7619 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7620 // CHECK11-NEXT:    ret i32 [[TMP8]]
7621 //
7622 //
7623 // CHECK13-LABEL: define {{[^@]+}}@_Z7get_valv
7624 // CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
7625 // CHECK13-NEXT:  entry:
7626 // CHECK13-NEXT:    ret i64 0
7627 //
7628 //
7629 // CHECK13-LABEL: define {{[^@]+}}@_Z3fooi
7630 // CHECK13-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
7631 // CHECK13-NEXT:  entry:
7632 // CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7633 // CHECK13-NEXT:    [[A:%.*]] = alloca i32, align 4
7634 // CHECK13-NEXT:    [[AA:%.*]] = alloca i16, align 2
7635 // CHECK13-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
7636 // CHECK13-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7637 // CHECK13-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7638 // CHECK13-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
7639 // CHECK13-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i64, align 8
7640 // CHECK13-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 8
7641 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7642 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7643 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7644 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7645 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
7646 // CHECK13-NEXT:    [[K:%.*]] = alloca i64, align 8
7647 // CHECK13-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
7648 // CHECK13-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
7649 // CHECK13-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
7650 // CHECK13-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
7651 // CHECK13-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
7652 // CHECK13-NEXT:    [[I7:%.*]] = alloca i32, align 4
7653 // CHECK13-NEXT:    [[K8:%.*]] = alloca i64, align 8
7654 // CHECK13-NEXT:    [[LIN:%.*]] = alloca i32, align 4
7655 // CHECK13-NEXT:    [[_TMP20:%.*]] = alloca i64, align 8
7656 // CHECK13-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i64, align 8
7657 // CHECK13-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i64, align 8
7658 // CHECK13-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i64, align 8
7659 // CHECK13-NEXT:    [[DOTLINEAR_START24:%.*]] = alloca i32, align 4
7660 // CHECK13-NEXT:    [[DOTLINEAR_START25:%.*]] = alloca i32, align 4
7661 // CHECK13-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
7662 // CHECK13-NEXT:    [[IT:%.*]] = alloca i64, align 8
7663 // CHECK13-NEXT:    [[LIN27:%.*]] = alloca i32, align 4
7664 // CHECK13-NEXT:    [[A28:%.*]] = alloca i32, align 4
7665 // CHECK13-NEXT:    [[_TMP49:%.*]] = alloca i16, align 2
7666 // CHECK13-NEXT:    [[DOTOMP_LB50:%.*]] = alloca i32, align 4
7667 // CHECK13-NEXT:    [[DOTOMP_UB51:%.*]] = alloca i32, align 4
7668 // CHECK13-NEXT:    [[DOTOMP_IV52:%.*]] = alloca i32, align 4
7669 // CHECK13-NEXT:    [[IT53:%.*]] = alloca i16, align 2
7670 // CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7671 // CHECK13-NEXT:    [[_TMP68:%.*]] = alloca i8, align 1
7672 // CHECK13-NEXT:    [[DOTOMP_LB69:%.*]] = alloca i32, align 4
7673 // CHECK13-NEXT:    [[DOTOMP_UB70:%.*]] = alloca i32, align 4
7674 // CHECK13-NEXT:    [[DOTOMP_IV71:%.*]] = alloca i32, align 4
7675 // CHECK13-NEXT:    [[IT72:%.*]] = alloca i8, align 1
7676 // CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7677 // CHECK13-NEXT:    store i32 0, i32* [[A]], align 4
7678 // CHECK13-NEXT:    store i16 0, i16* [[AA]], align 2
7679 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7680 // CHECK13-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
7681 // CHECK13-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
7682 // CHECK13-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
7683 // CHECK13-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 4
7684 // CHECK13-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
7685 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
7686 // CHECK13-NEXT:    [[TMP4:%.*]] = zext i32 [[TMP3]] to i64
7687 // CHECK13-NEXT:    [[TMP5:%.*]] = mul nuw i64 5, [[TMP4]]
7688 // CHECK13-NEXT:    [[VLA1:%.*]] = alloca double, i64 [[TMP5]], align 8
7689 // CHECK13-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR1]], align 8
7690 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7691 // CHECK13-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
7692 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7693 // CHECK13-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7694 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7695 // CHECK13:       omp.inner.for.cond:
7696 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7697 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
7698 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7699 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7700 // CHECK13:       omp.inner.for.body:
7701 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7702 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 5
7703 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
7704 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
7705 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7706 // CHECK13:       omp.body.continue:
7707 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7708 // CHECK13:       omp.inner.for.inc:
7709 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7710 // CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
7711 // CHECK13-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7712 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
7713 // CHECK13:       omp.inner.for.end:
7714 // CHECK13-NEXT:    store i32 33, i32* [[I]], align 4
7715 // CHECK13-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
7716 // CHECK13-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
7717 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
7718 // CHECK13-NEXT:    store i32 8, i32* [[DOTOMP_UB5]], align 4
7719 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
7720 // CHECK13-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV6]], align 4
7721 // CHECK13-NEXT:    [[TMP12:%.*]] = load i64, i64* [[K]], align 8
7722 // CHECK13-NEXT:    store i64 [[TMP12]], i64* [[DOTLINEAR_START]], align 8
7723 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND9:%.*]]
7724 // CHECK13:       omp.inner.for.cond9:
7725 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
7726 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
7727 // CHECK13-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
7728 // CHECK13-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
7729 // CHECK13:       omp.inner.for.body11:
7730 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
7731 // CHECK13-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[TMP15]], 1
7732 // CHECK13-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL12]]
7733 // CHECK13-NEXT:    store i32 [[SUB]], i32* [[I7]], align 4, !llvm.access.group !6
7734 // CHECK13-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !6
7735 // CHECK13-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
7736 // CHECK13-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP17]], 3
7737 // CHECK13-NEXT:    [[CONV:%.*]] = sext i32 [[MUL13]] to i64
7738 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i64 [[TMP16]], [[CONV]]
7739 // CHECK13-NEXT:    store i64 [[ADD14]], i64* [[K8]], align 8, !llvm.access.group !6
7740 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !6
7741 // CHECK13-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP18]], 1
7742 // CHECK13-NEXT:    store i32 [[ADD15]], i32* [[A]], align 4, !llvm.access.group !6
7743 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
7744 // CHECK13:       omp.body.continue16:
7745 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
7746 // CHECK13:       omp.inner.for.inc17:
7747 // CHECK13-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
7748 // CHECK13-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP19]], 1
7749 // CHECK13-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
7750 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP7:![0-9]+]]
7751 // CHECK13:       omp.inner.for.end19:
7752 // CHECK13-NEXT:    store i32 1, i32* [[I7]], align 4
7753 // CHECK13-NEXT:    [[TMP20:%.*]] = load i64, i64* [[K8]], align 8
7754 // CHECK13-NEXT:    store i64 [[TMP20]], i64* [[K]], align 8
7755 // CHECK13-NEXT:    store i32 12, i32* [[LIN]], align 4
7756 // CHECK13-NEXT:    store i64 0, i64* [[DOTOMP_LB21]], align 8
7757 // CHECK13-NEXT:    store i64 3, i64* [[DOTOMP_UB22]], align 8
7758 // CHECK13-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB21]], align 8
7759 // CHECK13-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV23]], align 8
7760 // CHECK13-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN]], align 4
7761 // CHECK13-NEXT:    store i32 [[TMP22]], i32* [[DOTLINEAR_START24]], align 4
7762 // CHECK13-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A]], align 4
7763 // CHECK13-NEXT:    store i32 [[TMP23]], i32* [[DOTLINEAR_START25]], align 4
7764 // CHECK13-NEXT:    [[CALL26:%.*]] = call noundef i64 @_Z7get_valv()
7765 // CHECK13-NEXT:    store i64 [[CALL26]], i64* [[DOTLINEAR_STEP]], align 8
7766 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
7767 // CHECK13:       omp.inner.for.cond29:
7768 // CHECK13-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7769 // CHECK13-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_UB22]], align 8, !llvm.access.group !9
7770 // CHECK13-NEXT:    [[CMP30:%.*]] = icmp ule i64 [[TMP24]], [[TMP25]]
7771 // CHECK13-NEXT:    br i1 [[CMP30]], label [[OMP_INNER_FOR_BODY31:%.*]], label [[OMP_INNER_FOR_END48:%.*]]
7772 // CHECK13:       omp.inner.for.body31:
7773 // CHECK13-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7774 // CHECK13-NEXT:    [[MUL32:%.*]] = mul i64 [[TMP26]], 400
7775 // CHECK13-NEXT:    [[SUB33:%.*]] = sub i64 2000, [[MUL32]]
7776 // CHECK13-NEXT:    store i64 [[SUB33]], i64* [[IT]], align 8, !llvm.access.group !9
7777 // CHECK13-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTLINEAR_START24]], align 4, !llvm.access.group !9
7778 // CHECK13-NEXT:    [[CONV34:%.*]] = sext i32 [[TMP27]] to i64
7779 // CHECK13-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7780 // CHECK13-NEXT:    [[TMP29:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !9
7781 // CHECK13-NEXT:    [[MUL35:%.*]] = mul i64 [[TMP28]], [[TMP29]]
7782 // CHECK13-NEXT:    [[ADD36:%.*]] = add i64 [[CONV34]], [[MUL35]]
7783 // CHECK13-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
7784 // CHECK13-NEXT:    store i32 [[CONV37]], i32* [[LIN27]], align 4, !llvm.access.group !9
7785 // CHECK13-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTLINEAR_START25]], align 4, !llvm.access.group !9
7786 // CHECK13-NEXT:    [[CONV38:%.*]] = sext i32 [[TMP30]] to i64
7787 // CHECK13-NEXT:    [[TMP31:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7788 // CHECK13-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !9
7789 // CHECK13-NEXT:    [[MUL39:%.*]] = mul i64 [[TMP31]], [[TMP32]]
7790 // CHECK13-NEXT:    [[ADD40:%.*]] = add i64 [[CONV38]], [[MUL39]]
7791 // CHECK13-NEXT:    [[CONV41:%.*]] = trunc i64 [[ADD40]] to i32
7792 // CHECK13-NEXT:    store i32 [[CONV41]], i32* [[A28]], align 4, !llvm.access.group !9
7793 // CHECK13-NEXT:    [[TMP33:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !9
7794 // CHECK13-NEXT:    [[CONV42:%.*]] = sext i16 [[TMP33]] to i32
7795 // CHECK13-NEXT:    [[ADD43:%.*]] = add nsw i32 [[CONV42]], 1
7796 // CHECK13-NEXT:    [[CONV44:%.*]] = trunc i32 [[ADD43]] to i16
7797 // CHECK13-NEXT:    store i16 [[CONV44]], i16* [[AA]], align 2, !llvm.access.group !9
7798 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE45:%.*]]
7799 // CHECK13:       omp.body.continue45:
7800 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC46:%.*]]
7801 // CHECK13:       omp.inner.for.inc46:
7802 // CHECK13-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7803 // CHECK13-NEXT:    [[ADD47:%.*]] = add i64 [[TMP34]], 1
7804 // CHECK13-NEXT:    store i64 [[ADD47]], i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !9
7805 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP10:![0-9]+]]
7806 // CHECK13:       omp.inner.for.end48:
7807 // CHECK13-NEXT:    store i64 400, i64* [[IT]], align 8
7808 // CHECK13-NEXT:    [[TMP35:%.*]] = load i32, i32* [[LIN27]], align 4
7809 // CHECK13-NEXT:    store i32 [[TMP35]], i32* [[LIN]], align 4
7810 // CHECK13-NEXT:    [[TMP36:%.*]] = load i32, i32* [[A28]], align 4
7811 // CHECK13-NEXT:    store i32 [[TMP36]], i32* [[A]], align 4
7812 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB50]], align 4
7813 // CHECK13-NEXT:    store i32 3, i32* [[DOTOMP_UB51]], align 4
7814 // CHECK13-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_LB50]], align 4
7815 // CHECK13-NEXT:    store i32 [[TMP37]], i32* [[DOTOMP_IV52]], align 4
7816 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND54:%.*]]
7817 // CHECK13:       omp.inner.for.cond54:
7818 // CHECK13-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
7819 // CHECK13-NEXT:    [[TMP39:%.*]] = load i32, i32* [[DOTOMP_UB51]], align 4, !llvm.access.group !12
7820 // CHECK13-NEXT:    [[CMP55:%.*]] = icmp sle i32 [[TMP38]], [[TMP39]]
7821 // CHECK13-NEXT:    br i1 [[CMP55]], label [[OMP_INNER_FOR_BODY56:%.*]], label [[OMP_INNER_FOR_END67:%.*]]
7822 // CHECK13:       omp.inner.for.body56:
7823 // CHECK13-NEXT:    [[TMP40:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
7824 // CHECK13-NEXT:    [[MUL57:%.*]] = mul nsw i32 [[TMP40]], 4
7825 // CHECK13-NEXT:    [[ADD58:%.*]] = add nsw i32 6, [[MUL57]]
7826 // CHECK13-NEXT:    [[CONV59:%.*]] = trunc i32 [[ADD58]] to i16
7827 // CHECK13-NEXT:    store i16 [[CONV59]], i16* [[IT53]], align 2, !llvm.access.group !12
7828 // CHECK13-NEXT:    [[TMP41:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !12
7829 // CHECK13-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP41]], 1
7830 // CHECK13-NEXT:    store i32 [[ADD60]], i32* [[A]], align 4, !llvm.access.group !12
7831 // CHECK13-NEXT:    [[TMP42:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !12
7832 // CHECK13-NEXT:    [[CONV61:%.*]] = sext i16 [[TMP42]] to i32
7833 // CHECK13-NEXT:    [[ADD62:%.*]] = add nsw i32 [[CONV61]], 1
7834 // CHECK13-NEXT:    [[CONV63:%.*]] = trunc i32 [[ADD62]] to i16
7835 // CHECK13-NEXT:    store i16 [[CONV63]], i16* [[AA]], align 2, !llvm.access.group !12
7836 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE64:%.*]]
7837 // CHECK13:       omp.body.continue64:
7838 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC65:%.*]]
7839 // CHECK13:       omp.inner.for.inc65:
7840 // CHECK13-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
7841 // CHECK13-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP43]], 1
7842 // CHECK13-NEXT:    store i32 [[ADD66]], i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !12
7843 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND54]], !llvm.loop [[LOOP13:![0-9]+]]
7844 // CHECK13:       omp.inner.for.end67:
7845 // CHECK13-NEXT:    store i16 22, i16* [[IT53]], align 2
7846 // CHECK13-NEXT:    [[TMP44:%.*]] = load i32, i32* [[A]], align 4
7847 // CHECK13-NEXT:    store i32 [[TMP44]], i32* [[DOTCAPTURE_EXPR_]], align 4
7848 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB69]], align 4
7849 // CHECK13-NEXT:    store i32 25, i32* [[DOTOMP_UB70]], align 4
7850 // CHECK13-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_LB69]], align 4
7851 // CHECK13-NEXT:    store i32 [[TMP45]], i32* [[DOTOMP_IV71]], align 4
7852 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND73:%.*]]
7853 // CHECK13:       omp.inner.for.cond73:
7854 // CHECK13-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
7855 // CHECK13-NEXT:    [[TMP47:%.*]] = load i32, i32* [[DOTOMP_UB70]], align 4, !llvm.access.group !15
7856 // CHECK13-NEXT:    [[CMP74:%.*]] = icmp sle i32 [[TMP46]], [[TMP47]]
7857 // CHECK13-NEXT:    br i1 [[CMP74]], label [[OMP_INNER_FOR_BODY75:%.*]], label [[OMP_INNER_FOR_END100:%.*]]
7858 // CHECK13:       omp.inner.for.body75:
7859 // CHECK13-NEXT:    [[TMP48:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
7860 // CHECK13-NEXT:    [[MUL76:%.*]] = mul nsw i32 [[TMP48]], 1
7861 // CHECK13-NEXT:    [[SUB77:%.*]] = sub nsw i32 122, [[MUL76]]
7862 // CHECK13-NEXT:    [[CONV78:%.*]] = trunc i32 [[SUB77]] to i8
7863 // CHECK13-NEXT:    store i8 [[CONV78]], i8* [[IT72]], align 1, !llvm.access.group !15
7864 // CHECK13-NEXT:    [[TMP49:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !15
7865 // CHECK13-NEXT:    [[ADD79:%.*]] = add nsw i32 [[TMP49]], 1
7866 // CHECK13-NEXT:    store i32 [[ADD79]], i32* [[A]], align 4, !llvm.access.group !15
7867 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[B]], i64 0, i64 2
7868 // CHECK13-NEXT:    [[TMP50:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !15
7869 // CHECK13-NEXT:    [[CONV80:%.*]] = fpext float [[TMP50]] to double
7870 // CHECK13-NEXT:    [[ADD81:%.*]] = fadd double [[CONV80]], 1.000000e+00
7871 // CHECK13-NEXT:    [[CONV82:%.*]] = fptrunc double [[ADD81]] to float
7872 // CHECK13-NEXT:    store float [[CONV82]], float* [[ARRAYIDX]], align 4, !llvm.access.group !15
7873 // CHECK13-NEXT:    [[ARRAYIDX83:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 3
7874 // CHECK13-NEXT:    [[TMP51:%.*]] = load float, float* [[ARRAYIDX83]], align 4, !llvm.access.group !15
7875 // CHECK13-NEXT:    [[CONV84:%.*]] = fpext float [[TMP51]] to double
7876 // CHECK13-NEXT:    [[ADD85:%.*]] = fadd double [[CONV84]], 1.000000e+00
7877 // CHECK13-NEXT:    [[CONV86:%.*]] = fptrunc double [[ADD85]] to float
7878 // CHECK13-NEXT:    store float [[CONV86]], float* [[ARRAYIDX83]], align 4, !llvm.access.group !15
7879 // CHECK13-NEXT:    [[ARRAYIDX87:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[C]], i64 0, i64 1
7880 // CHECK13-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX87]], i64 0, i64 2
7881 // CHECK13-NEXT:    [[TMP52:%.*]] = load double, double* [[ARRAYIDX88]], align 8, !llvm.access.group !15
7882 // CHECK13-NEXT:    [[ADD89:%.*]] = fadd double [[TMP52]], 1.000000e+00
7883 // CHECK13-NEXT:    store double [[ADD89]], double* [[ARRAYIDX88]], align 8, !llvm.access.group !15
7884 // CHECK13-NEXT:    [[TMP53:%.*]] = mul nsw i64 1, [[TMP4]]
7885 // CHECK13-NEXT:    [[ARRAYIDX90:%.*]] = getelementptr inbounds double, double* [[VLA1]], i64 [[TMP53]]
7886 // CHECK13-NEXT:    [[ARRAYIDX91:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX90]], i64 3
7887 // CHECK13-NEXT:    [[TMP54:%.*]] = load double, double* [[ARRAYIDX91]], align 8, !llvm.access.group !15
7888 // CHECK13-NEXT:    [[ADD92:%.*]] = fadd double [[TMP54]], 1.000000e+00
7889 // CHECK13-NEXT:    store double [[ADD92]], double* [[ARRAYIDX91]], align 8, !llvm.access.group !15
7890 // CHECK13-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 0
7891 // CHECK13-NEXT:    [[TMP55:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !15
7892 // CHECK13-NEXT:    [[ADD93:%.*]] = add nsw i64 [[TMP55]], 1
7893 // CHECK13-NEXT:    store i64 [[ADD93]], i64* [[X]], align 8, !llvm.access.group !15
7894 // CHECK13-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 1
7895 // CHECK13-NEXT:    [[TMP56:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !15
7896 // CHECK13-NEXT:    [[CONV94:%.*]] = sext i8 [[TMP56]] to i32
7897 // CHECK13-NEXT:    [[ADD95:%.*]] = add nsw i32 [[CONV94]], 1
7898 // CHECK13-NEXT:    [[CONV96:%.*]] = trunc i32 [[ADD95]] to i8
7899 // CHECK13-NEXT:    store i8 [[CONV96]], i8* [[Y]], align 8, !llvm.access.group !15
7900 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE97:%.*]]
7901 // CHECK13:       omp.body.continue97:
7902 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC98:%.*]]
7903 // CHECK13:       omp.inner.for.inc98:
7904 // CHECK13-NEXT:    [[TMP57:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
7905 // CHECK13-NEXT:    [[ADD99:%.*]] = add nsw i32 [[TMP57]], 1
7906 // CHECK13-NEXT:    store i32 [[ADD99]], i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !15
7907 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND73]], !llvm.loop [[LOOP16:![0-9]+]]
7908 // CHECK13:       omp.inner.for.end100:
7909 // CHECK13-NEXT:    store i8 96, i8* [[IT72]], align 1
7910 // CHECK13-NEXT:    [[TMP58:%.*]] = load i32, i32* [[A]], align 4
7911 // CHECK13-NEXT:    [[TMP59:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7912 // CHECK13-NEXT:    call void @llvm.stackrestore(i8* [[TMP59]])
7913 // CHECK13-NEXT:    ret i32 [[TMP58]]
7914 //
7915 //
7916 // CHECK13-LABEL: define {{[^@]+}}@_Z3bari
7917 // CHECK13-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
7918 // CHECK13-NEXT:  entry:
7919 // CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7920 // CHECK13-NEXT:    [[A:%.*]] = alloca i32, align 4
7921 // CHECK13-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 8
7922 // CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7923 // CHECK13-NEXT:    store i32 0, i32* [[A]], align 4
7924 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7925 // CHECK13-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z3fooi(i32 noundef signext [[TMP0]])
7926 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
7927 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
7928 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
7929 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
7930 // CHECK13-NEXT:    [[CALL1:%.*]] = call noundef signext i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 8 dereferenceable(8) [[S]], i32 noundef signext [[TMP2]])
7931 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
7932 // CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
7933 // CHECK13-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
7934 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
7935 // CHECK13-NEXT:    [[CALL3:%.*]] = call noundef signext i32 @_ZL7fstatici(i32 noundef signext [[TMP4]])
7936 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
7937 // CHECK13-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
7938 // CHECK13-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
7939 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
7940 // CHECK13-NEXT:    [[CALL5:%.*]] = call noundef signext i32 @_Z9ftemplateIiET_i(i32 noundef signext [[TMP6]])
7941 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
7942 // CHECK13-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
7943 // CHECK13-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
7944 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
7945 // CHECK13-NEXT:    ret i32 [[TMP8]]
7946 //
7947 //
7948 // CHECK13-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
7949 // CHECK13-SAME: (%struct.S1* noundef nonnull align 8 dereferenceable(8) [[THIS:%.*]], i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat align 2 {
7950 // CHECK13-NEXT:  entry:
7951 // CHECK13-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
7952 // CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7953 // CHECK13-NEXT:    [[B:%.*]] = alloca i32, align 4
7954 // CHECK13-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7955 // CHECK13-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7956 // CHECK13-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
7957 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i64, align 8
7958 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7959 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7960 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7961 // CHECK13-NEXT:    [[IT:%.*]] = alloca i64, align 8
7962 // CHECK13-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
7963 // CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7964 // CHECK13-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
7965 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7966 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
7967 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
7968 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
7969 // CHECK13-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
7970 // CHECK13-NEXT:    [[TMP3:%.*]] = call i8* @llvm.stacksave()
7971 // CHECK13-NEXT:    store i8* [[TMP3]], i8** [[SAVED_STACK]], align 8
7972 // CHECK13-NEXT:    [[TMP4:%.*]] = mul nuw i64 2, [[TMP2]]
7973 // CHECK13-NEXT:    [[VLA:%.*]] = alloca i16, i64 [[TMP4]], align 2
7974 // CHECK13-NEXT:    store i64 [[TMP2]], i64* [[__VLA_EXPR0]], align 8
7975 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
7976 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 60
7977 // CHECK13-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
7978 // CHECK13-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
7979 // CHECK13-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7980 // CHECK13-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
7981 // CHECK13-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7982 // CHECK13-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
7983 // CHECK13-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
7984 // CHECK13-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
7985 // CHECK13-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7986 // CHECK13:       omp_if.then:
7987 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7988 // CHECK13:       omp.inner.for.cond:
7989 // CHECK13-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7990 // CHECK13-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !18
7991 // CHECK13-NEXT:    [[CMP2:%.*]] = icmp ule i64 [[TMP8]], [[TMP9]]
7992 // CHECK13-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7993 // CHECK13:       omp.inner.for.body:
7994 // CHECK13-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
7995 // CHECK13-NEXT:    [[MUL:%.*]] = mul i64 [[TMP10]], 400
7996 // CHECK13-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
7997 // CHECK13-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !18
7998 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[B]], align 4, !llvm.access.group !18
7999 // CHECK13-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP11]] to double
8000 // CHECK13-NEXT:    [[ADD3:%.*]] = fadd double [[CONV]], 1.500000e+00
8001 // CHECK13-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
8002 // CHECK13-NEXT:    store double [[ADD3]], double* [[A]], align 8, !nontemporal !19, !llvm.access.group !18
8003 // CHECK13-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8004 // CHECK13-NEXT:    [[TMP12:%.*]] = load double, double* [[A4]], align 8, !nontemporal !19, !llvm.access.group !18
8005 // CHECK13-NEXT:    [[INC:%.*]] = fadd double [[TMP12]], 1.000000e+00
8006 // CHECK13-NEXT:    store double [[INC]], double* [[A4]], align 8, !nontemporal !19, !llvm.access.group !18
8007 // CHECK13-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
8008 // CHECK13-NEXT:    [[TMP13:%.*]] = mul nsw i64 1, [[TMP2]]
8009 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP13]]
8010 // CHECK13-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
8011 // CHECK13-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2, !llvm.access.group !18
8012 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8013 // CHECK13:       omp.body.continue:
8014 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8015 // CHECK13:       omp.inner.for.inc:
8016 // CHECK13-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
8017 // CHECK13-NEXT:    [[ADD7:%.*]] = add i64 [[TMP14]], 1
8018 // CHECK13-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
8019 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
8020 // CHECK13:       omp.inner.for.end:
8021 // CHECK13-NEXT:    br label [[OMP_IF_END:%.*]]
8022 // CHECK13:       omp_if.else:
8023 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
8024 // CHECK13:       omp.inner.for.cond8:
8025 // CHECK13-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8026 // CHECK13-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8027 // CHECK13-NEXT:    [[CMP9:%.*]] = icmp ule i64 [[TMP15]], [[TMP16]]
8028 // CHECK13-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END24:%.*]]
8029 // CHECK13:       omp.inner.for.body10:
8030 // CHECK13-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8031 // CHECK13-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP17]], 400
8032 // CHECK13-NEXT:    [[SUB12:%.*]] = sub i64 2000, [[MUL11]]
8033 // CHECK13-NEXT:    store i64 [[SUB12]], i64* [[IT]], align 8
8034 // CHECK13-NEXT:    [[TMP18:%.*]] = load i32, i32* [[B]], align 4
8035 // CHECK13-NEXT:    [[CONV13:%.*]] = sitofp i32 [[TMP18]] to double
8036 // CHECK13-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.500000e+00
8037 // CHECK13-NEXT:    [[A15:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8038 // CHECK13-NEXT:    store double [[ADD14]], double* [[A15]], align 8
8039 // CHECK13-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8040 // CHECK13-NEXT:    [[TMP19:%.*]] = load double, double* [[A16]], align 8
8041 // CHECK13-NEXT:    [[INC17:%.*]] = fadd double [[TMP19]], 1.000000e+00
8042 // CHECK13-NEXT:    store double [[INC17]], double* [[A16]], align 8
8043 // CHECK13-NEXT:    [[CONV18:%.*]] = fptosi double [[INC17]] to i16
8044 // CHECK13-NEXT:    [[TMP20:%.*]] = mul nsw i64 1, [[TMP2]]
8045 // CHECK13-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP20]]
8046 // CHECK13-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX19]], i64 1
8047 // CHECK13-NEXT:    store i16 [[CONV18]], i16* [[ARRAYIDX20]], align 2
8048 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE21:%.*]]
8049 // CHECK13:       omp.body.continue21:
8050 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC22:%.*]]
8051 // CHECK13:       omp.inner.for.inc22:
8052 // CHECK13-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8053 // CHECK13-NEXT:    [[ADD23:%.*]] = add i64 [[TMP21]], 1
8054 // CHECK13-NEXT:    store i64 [[ADD23]], i64* [[DOTOMP_IV]], align 8
8055 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP22:![0-9]+]]
8056 // CHECK13:       omp.inner.for.end24:
8057 // CHECK13-NEXT:    br label [[OMP_IF_END]]
8058 // CHECK13:       omp_if.end:
8059 // CHECK13-NEXT:    store i64 400, i64* [[IT]], align 8
8060 // CHECK13-NEXT:    [[TMP22:%.*]] = mul nsw i64 1, [[TMP2]]
8061 // CHECK13-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i64 [[TMP22]]
8062 // CHECK13-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX25]], i64 1
8063 // CHECK13-NEXT:    [[TMP23:%.*]] = load i16, i16* [[ARRAYIDX26]], align 2
8064 // CHECK13-NEXT:    [[CONV27:%.*]] = sext i16 [[TMP23]] to i32
8065 // CHECK13-NEXT:    [[TMP24:%.*]] = load i32, i32* [[B]], align 4
8066 // CHECK13-NEXT:    [[ADD28:%.*]] = add nsw i32 [[CONV27]], [[TMP24]]
8067 // CHECK13-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
8068 // CHECK13-NEXT:    call void @llvm.stackrestore(i8* [[TMP25]])
8069 // CHECK13-NEXT:    ret i32 [[ADD28]]
8070 //
8071 //
8072 // CHECK13-LABEL: define {{[^@]+}}@_ZL7fstatici
8073 // CHECK13-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] {
8074 // CHECK13-NEXT:  entry:
8075 // CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8076 // CHECK13-NEXT:    [[A:%.*]] = alloca i32, align 4
8077 // CHECK13-NEXT:    [[AA:%.*]] = alloca i16, align 2
8078 // CHECK13-NEXT:    [[AAA:%.*]] = alloca i8, align 1
8079 // CHECK13-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8080 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8081 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8082 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8083 // CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8084 // CHECK13-NEXT:    store i32 0, i32* [[A]], align 4
8085 // CHECK13-NEXT:    store i16 0, i16* [[AA]], align 2
8086 // CHECK13-NEXT:    store i8 0, i8* [[AAA]], align 1
8087 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8088 // CHECK13-NEXT:    store i32 429496720, i32* [[DOTOMP_UB]], align 4
8089 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8090 // CHECK13-NEXT:    ret i32 [[TMP0]]
8091 //
8092 //
8093 // CHECK13-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
8094 // CHECK13-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0]] comdat {
8095 // CHECK13-NEXT:  entry:
8096 // CHECK13-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8097 // CHECK13-NEXT:    [[A:%.*]] = alloca i32, align 4
8098 // CHECK13-NEXT:    [[AA:%.*]] = alloca i16, align 2
8099 // CHECK13-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8100 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i64, align 8
8101 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
8102 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
8103 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
8104 // CHECK13-NEXT:    [[I:%.*]] = alloca i64, align 8
8105 // CHECK13-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8106 // CHECK13-NEXT:    store i32 0, i32* [[A]], align 4
8107 // CHECK13-NEXT:    store i16 0, i16* [[AA]], align 2
8108 // CHECK13-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
8109 // CHECK13-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
8110 // CHECK13-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
8111 // CHECK13-NEXT:    store i64 [[TMP0]], i64* [[DOTOMP_IV]], align 8
8112 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8113 // CHECK13:       omp.inner.for.cond:
8114 // CHECK13-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !24
8115 // CHECK13-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !24
8116 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i64 [[TMP1]], [[TMP2]]
8117 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8118 // CHECK13:       omp.inner.for.body:
8119 // CHECK13-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !24
8120 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP3]], 3
8121 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
8122 // CHECK13-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !24
8123 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !24
8124 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
8125 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[A]], align 4, !llvm.access.group !24
8126 // CHECK13-NEXT:    [[TMP5:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !24
8127 // CHECK13-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
8128 // CHECK13-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV]], 1
8129 // CHECK13-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
8130 // CHECK13-NEXT:    store i16 [[CONV3]], i16* [[AA]], align 2, !llvm.access.group !24
8131 // CHECK13-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[B]], i64 0, i64 2
8132 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
8133 // CHECK13-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP6]], 1
8134 // CHECK13-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
8135 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8136 // CHECK13:       omp.body.continue:
8137 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8138 // CHECK13:       omp.inner.for.inc:
8139 // CHECK13-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !24
8140 // CHECK13-NEXT:    [[ADD5:%.*]] = add nsw i64 [[TMP7]], 1
8141 // CHECK13-NEXT:    store i64 [[ADD5]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !24
8142 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
8143 // CHECK13:       omp.inner.for.end:
8144 // CHECK13-NEXT:    store i64 11, i64* [[I]], align 8
8145 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
8146 // CHECK13-NEXT:    ret i32 [[TMP8]]
8147 //
8148 //
8149 // CHECK15-LABEL: define {{[^@]+}}@_Z7get_valv
8150 // CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
8151 // CHECK15-NEXT:  entry:
8152 // CHECK15-NEXT:    ret i64 0
8153 //
8154 //
8155 // CHECK15-LABEL: define {{[^@]+}}@_Z3fooi
8156 // CHECK15-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
8157 // CHECK15-NEXT:  entry:
8158 // CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8159 // CHECK15-NEXT:    [[A:%.*]] = alloca i32, align 4
8160 // CHECK15-NEXT:    [[AA:%.*]] = alloca i16, align 2
8161 // CHECK15-NEXT:    [[B:%.*]] = alloca [10 x float], align 4
8162 // CHECK15-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
8163 // CHECK15-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
8164 // CHECK15-NEXT:    [[C:%.*]] = alloca [5 x [10 x double]], align 8
8165 // CHECK15-NEXT:    [[__VLA_EXPR1:%.*]] = alloca i32, align 4
8166 // CHECK15-NEXT:    [[D:%.*]] = alloca [[STRUCT_TT:%.*]], align 4
8167 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8168 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8169 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8170 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8171 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
8172 // CHECK15-NEXT:    [[K:%.*]] = alloca i64, align 8
8173 // CHECK15-NEXT:    [[_TMP3:%.*]] = alloca i32, align 4
8174 // CHECK15-NEXT:    [[DOTOMP_LB4:%.*]] = alloca i32, align 4
8175 // CHECK15-NEXT:    [[DOTOMP_UB5:%.*]] = alloca i32, align 4
8176 // CHECK15-NEXT:    [[DOTOMP_IV6:%.*]] = alloca i32, align 4
8177 // CHECK15-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i64, align 8
8178 // CHECK15-NEXT:    [[I7:%.*]] = alloca i32, align 4
8179 // CHECK15-NEXT:    [[K8:%.*]] = alloca i64, align 8
8180 // CHECK15-NEXT:    [[LIN:%.*]] = alloca i32, align 4
8181 // CHECK15-NEXT:    [[_TMP20:%.*]] = alloca i64, align 4
8182 // CHECK15-NEXT:    [[DOTOMP_LB21:%.*]] = alloca i64, align 8
8183 // CHECK15-NEXT:    [[DOTOMP_UB22:%.*]] = alloca i64, align 8
8184 // CHECK15-NEXT:    [[DOTOMP_IV23:%.*]] = alloca i64, align 8
8185 // CHECK15-NEXT:    [[DOTLINEAR_START24:%.*]] = alloca i32, align 4
8186 // CHECK15-NEXT:    [[DOTLINEAR_START25:%.*]] = alloca i32, align 4
8187 // CHECK15-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
8188 // CHECK15-NEXT:    [[IT:%.*]] = alloca i64, align 8
8189 // CHECK15-NEXT:    [[LIN27:%.*]] = alloca i32, align 4
8190 // CHECK15-NEXT:    [[A28:%.*]] = alloca i32, align 4
8191 // CHECK15-NEXT:    [[_TMP49:%.*]] = alloca i16, align 2
8192 // CHECK15-NEXT:    [[DOTOMP_LB50:%.*]] = alloca i32, align 4
8193 // CHECK15-NEXT:    [[DOTOMP_UB51:%.*]] = alloca i32, align 4
8194 // CHECK15-NEXT:    [[DOTOMP_IV52:%.*]] = alloca i32, align 4
8195 // CHECK15-NEXT:    [[IT53:%.*]] = alloca i16, align 2
8196 // CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8197 // CHECK15-NEXT:    [[_TMP68:%.*]] = alloca i8, align 1
8198 // CHECK15-NEXT:    [[DOTOMP_LB69:%.*]] = alloca i32, align 4
8199 // CHECK15-NEXT:    [[DOTOMP_UB70:%.*]] = alloca i32, align 4
8200 // CHECK15-NEXT:    [[DOTOMP_IV71:%.*]] = alloca i32, align 4
8201 // CHECK15-NEXT:    [[IT72:%.*]] = alloca i8, align 1
8202 // CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8203 // CHECK15-NEXT:    store i32 0, i32* [[A]], align 4
8204 // CHECK15-NEXT:    store i16 0, i16* [[AA]], align 2
8205 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8206 // CHECK15-NEXT:    [[TMP1:%.*]] = call i8* @llvm.stacksave()
8207 // CHECK15-NEXT:    store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
8208 // CHECK15-NEXT:    [[VLA:%.*]] = alloca float, i32 [[TMP0]], align 4
8209 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
8210 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
8211 // CHECK15-NEXT:    [[TMP3:%.*]] = mul nuw i32 5, [[TMP2]]
8212 // CHECK15-NEXT:    [[VLA1:%.*]] = alloca double, i32 [[TMP3]], align 8
8213 // CHECK15-NEXT:    store i32 [[TMP2]], i32* [[__VLA_EXPR1]], align 4
8214 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8215 // CHECK15-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
8216 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8217 // CHECK15-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8218 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8219 // CHECK15:       omp.inner.for.cond:
8220 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
8221 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
8222 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8223 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8224 // CHECK15:       omp.inner.for.body:
8225 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
8226 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
8227 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
8228 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
8229 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8230 // CHECK15:       omp.body.continue:
8231 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8232 // CHECK15:       omp.inner.for.inc:
8233 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
8234 // CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
8235 // CHECK15-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
8236 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
8237 // CHECK15:       omp.inner.for.end:
8238 // CHECK15-NEXT:    store i32 33, i32* [[I]], align 4
8239 // CHECK15-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv()
8240 // CHECK15-NEXT:    store i64 [[CALL]], i64* [[K]], align 8
8241 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB4]], align 4
8242 // CHECK15-NEXT:    store i32 8, i32* [[DOTOMP_UB5]], align 4
8243 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
8244 // CHECK15-NEXT:    store i32 [[TMP9]], i32* [[DOTOMP_IV6]], align 4
8245 // CHECK15-NEXT:    [[TMP10:%.*]] = load i64, i64* [[K]], align 8
8246 // CHECK15-NEXT:    store i64 [[TMP10]], i64* [[DOTLINEAR_START]], align 8
8247 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND9:%.*]]
8248 // CHECK15:       omp.inner.for.cond9:
8249 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
8250 // CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
8251 // CHECK15-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
8252 // CHECK15-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY11:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
8253 // CHECK15:       omp.inner.for.body11:
8254 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
8255 // CHECK15-NEXT:    [[MUL12:%.*]] = mul nsw i32 [[TMP13]], 1
8256 // CHECK15-NEXT:    [[SUB:%.*]] = sub nsw i32 10, [[MUL12]]
8257 // CHECK15-NEXT:    store i32 [[SUB]], i32* [[I7]], align 4, !llvm.access.group !7
8258 // CHECK15-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTLINEAR_START]], align 8, !llvm.access.group !7
8259 // CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
8260 // CHECK15-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP15]], 3
8261 // CHECK15-NEXT:    [[CONV:%.*]] = sext i32 [[MUL13]] to i64
8262 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i64 [[TMP14]], [[CONV]]
8263 // CHECK15-NEXT:    store i64 [[ADD14]], i64* [[K8]], align 8, !llvm.access.group !7
8264 // CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !7
8265 // CHECK15-NEXT:    [[ADD15:%.*]] = add nsw i32 [[TMP16]], 1
8266 // CHECK15-NEXT:    store i32 [[ADD15]], i32* [[A]], align 4, !llvm.access.group !7
8267 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE16:%.*]]
8268 // CHECK15:       omp.body.continue16:
8269 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC17:%.*]]
8270 // CHECK15:       omp.inner.for.inc17:
8271 // CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
8272 // CHECK15-NEXT:    [[ADD18:%.*]] = add nsw i32 [[TMP17]], 1
8273 // CHECK15-NEXT:    store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
8274 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND9]], !llvm.loop [[LOOP8:![0-9]+]]
8275 // CHECK15:       omp.inner.for.end19:
8276 // CHECK15-NEXT:    store i32 1, i32* [[I7]], align 4
8277 // CHECK15-NEXT:    [[TMP18:%.*]] = load i64, i64* [[K8]], align 8
8278 // CHECK15-NEXT:    store i64 [[TMP18]], i64* [[K]], align 8
8279 // CHECK15-NEXT:    store i32 12, i32* [[LIN]], align 4
8280 // CHECK15-NEXT:    store i64 0, i64* [[DOTOMP_LB21]], align 8
8281 // CHECK15-NEXT:    store i64 3, i64* [[DOTOMP_UB22]], align 8
8282 // CHECK15-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_LB21]], align 8
8283 // CHECK15-NEXT:    store i64 [[TMP19]], i64* [[DOTOMP_IV23]], align 8
8284 // CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[LIN]], align 4
8285 // CHECK15-NEXT:    store i32 [[TMP20]], i32* [[DOTLINEAR_START24]], align 4
8286 // CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* [[A]], align 4
8287 // CHECK15-NEXT:    store i32 [[TMP21]], i32* [[DOTLINEAR_START25]], align 4
8288 // CHECK15-NEXT:    [[CALL26:%.*]] = call noundef i64 @_Z7get_valv()
8289 // CHECK15-NEXT:    store i64 [[CALL26]], i64* [[DOTLINEAR_STEP]], align 8
8290 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND29:%.*]]
8291 // CHECK15:       omp.inner.for.cond29:
8292 // CHECK15-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8293 // CHECK15-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB22]], align 8, !llvm.access.group !10
8294 // CHECK15-NEXT:    [[CMP30:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
8295 // CHECK15-NEXT:    br i1 [[CMP30]], label [[OMP_INNER_FOR_BODY31:%.*]], label [[OMP_INNER_FOR_END48:%.*]]
8296 // CHECK15:       omp.inner.for.body31:
8297 // CHECK15-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8298 // CHECK15-NEXT:    [[MUL32:%.*]] = mul i64 [[TMP24]], 400
8299 // CHECK15-NEXT:    [[SUB33:%.*]] = sub i64 2000, [[MUL32]]
8300 // CHECK15-NEXT:    store i64 [[SUB33]], i64* [[IT]], align 8, !llvm.access.group !10
8301 // CHECK15-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTLINEAR_START24]], align 4, !llvm.access.group !10
8302 // CHECK15-NEXT:    [[CONV34:%.*]] = sext i32 [[TMP25]] to i64
8303 // CHECK15-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8304 // CHECK15-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !10
8305 // CHECK15-NEXT:    [[MUL35:%.*]] = mul i64 [[TMP26]], [[TMP27]]
8306 // CHECK15-NEXT:    [[ADD36:%.*]] = add i64 [[CONV34]], [[MUL35]]
8307 // CHECK15-NEXT:    [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
8308 // CHECK15-NEXT:    store i32 [[CONV37]], i32* [[LIN27]], align 4, !llvm.access.group !10
8309 // CHECK15-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTLINEAR_START25]], align 4, !llvm.access.group !10
8310 // CHECK15-NEXT:    [[CONV38:%.*]] = sext i32 [[TMP28]] to i64
8311 // CHECK15-NEXT:    [[TMP29:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8312 // CHECK15-NEXT:    [[TMP30:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !10
8313 // CHECK15-NEXT:    [[MUL39:%.*]] = mul i64 [[TMP29]], [[TMP30]]
8314 // CHECK15-NEXT:    [[ADD40:%.*]] = add i64 [[CONV38]], [[MUL39]]
8315 // CHECK15-NEXT:    [[CONV41:%.*]] = trunc i64 [[ADD40]] to i32
8316 // CHECK15-NEXT:    store i32 [[CONV41]], i32* [[A28]], align 4, !llvm.access.group !10
8317 // CHECK15-NEXT:    [[TMP31:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !10
8318 // CHECK15-NEXT:    [[CONV42:%.*]] = sext i16 [[TMP31]] to i32
8319 // CHECK15-NEXT:    [[ADD43:%.*]] = add nsw i32 [[CONV42]], 1
8320 // CHECK15-NEXT:    [[CONV44:%.*]] = trunc i32 [[ADD43]] to i16
8321 // CHECK15-NEXT:    store i16 [[CONV44]], i16* [[AA]], align 2, !llvm.access.group !10
8322 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE45:%.*]]
8323 // CHECK15:       omp.body.continue45:
8324 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC46:%.*]]
8325 // CHECK15:       omp.inner.for.inc46:
8326 // CHECK15-NEXT:    [[TMP32:%.*]] = load i64, i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8327 // CHECK15-NEXT:    [[ADD47:%.*]] = add i64 [[TMP32]], 1
8328 // CHECK15-NEXT:    store i64 [[ADD47]], i64* [[DOTOMP_IV23]], align 8, !llvm.access.group !10
8329 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND29]], !llvm.loop [[LOOP11:![0-9]+]]
8330 // CHECK15:       omp.inner.for.end48:
8331 // CHECK15-NEXT:    store i64 400, i64* [[IT]], align 8
8332 // CHECK15-NEXT:    [[TMP33:%.*]] = load i32, i32* [[LIN27]], align 4
8333 // CHECK15-NEXT:    store i32 [[TMP33]], i32* [[LIN]], align 4
8334 // CHECK15-NEXT:    [[TMP34:%.*]] = load i32, i32* [[A28]], align 4
8335 // CHECK15-NEXT:    store i32 [[TMP34]], i32* [[A]], align 4
8336 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB50]], align 4
8337 // CHECK15-NEXT:    store i32 3, i32* [[DOTOMP_UB51]], align 4
8338 // CHECK15-NEXT:    [[TMP35:%.*]] = load i32, i32* [[DOTOMP_LB50]], align 4
8339 // CHECK15-NEXT:    store i32 [[TMP35]], i32* [[DOTOMP_IV52]], align 4
8340 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND54:%.*]]
8341 // CHECK15:       omp.inner.for.cond54:
8342 // CHECK15-NEXT:    [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
8343 // CHECK15-NEXT:    [[TMP37:%.*]] = load i32, i32* [[DOTOMP_UB51]], align 4, !llvm.access.group !13
8344 // CHECK15-NEXT:    [[CMP55:%.*]] = icmp sle i32 [[TMP36]], [[TMP37]]
8345 // CHECK15-NEXT:    br i1 [[CMP55]], label [[OMP_INNER_FOR_BODY56:%.*]], label [[OMP_INNER_FOR_END67:%.*]]
8346 // CHECK15:       omp.inner.for.body56:
8347 // CHECK15-NEXT:    [[TMP38:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
8348 // CHECK15-NEXT:    [[MUL57:%.*]] = mul nsw i32 [[TMP38]], 4
8349 // CHECK15-NEXT:    [[ADD58:%.*]] = add nsw i32 6, [[MUL57]]
8350 // CHECK15-NEXT:    [[CONV59:%.*]] = trunc i32 [[ADD58]] to i16
8351 // CHECK15-NEXT:    store i16 [[CONV59]], i16* [[IT53]], align 2, !llvm.access.group !13
8352 // CHECK15-NEXT:    [[TMP39:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !13
8353 // CHECK15-NEXT:    [[ADD60:%.*]] = add nsw i32 [[TMP39]], 1
8354 // CHECK15-NEXT:    store i32 [[ADD60]], i32* [[A]], align 4, !llvm.access.group !13
8355 // CHECK15-NEXT:    [[TMP40:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !13
8356 // CHECK15-NEXT:    [[CONV61:%.*]] = sext i16 [[TMP40]] to i32
8357 // CHECK15-NEXT:    [[ADD62:%.*]] = add nsw i32 [[CONV61]], 1
8358 // CHECK15-NEXT:    [[CONV63:%.*]] = trunc i32 [[ADD62]] to i16
8359 // CHECK15-NEXT:    store i16 [[CONV63]], i16* [[AA]], align 2, !llvm.access.group !13
8360 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE64:%.*]]
8361 // CHECK15:       omp.body.continue64:
8362 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC65:%.*]]
8363 // CHECK15:       omp.inner.for.inc65:
8364 // CHECK15-NEXT:    [[TMP41:%.*]] = load i32, i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
8365 // CHECK15-NEXT:    [[ADD66:%.*]] = add nsw i32 [[TMP41]], 1
8366 // CHECK15-NEXT:    store i32 [[ADD66]], i32* [[DOTOMP_IV52]], align 4, !llvm.access.group !13
8367 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND54]], !llvm.loop [[LOOP14:![0-9]+]]
8368 // CHECK15:       omp.inner.for.end67:
8369 // CHECK15-NEXT:    store i16 22, i16* [[IT53]], align 2
8370 // CHECK15-NEXT:    [[TMP42:%.*]] = load i32, i32* [[A]], align 4
8371 // CHECK15-NEXT:    store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_]], align 4
8372 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB69]], align 4
8373 // CHECK15-NEXT:    store i32 25, i32* [[DOTOMP_UB70]], align 4
8374 // CHECK15-NEXT:    [[TMP43:%.*]] = load i32, i32* [[DOTOMP_LB69]], align 4
8375 // CHECK15-NEXT:    store i32 [[TMP43]], i32* [[DOTOMP_IV71]], align 4
8376 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND73:%.*]]
8377 // CHECK15:       omp.inner.for.cond73:
8378 // CHECK15-NEXT:    [[TMP44:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
8379 // CHECK15-NEXT:    [[TMP45:%.*]] = load i32, i32* [[DOTOMP_UB70]], align 4, !llvm.access.group !16
8380 // CHECK15-NEXT:    [[CMP74:%.*]] = icmp sle i32 [[TMP44]], [[TMP45]]
8381 // CHECK15-NEXT:    br i1 [[CMP74]], label [[OMP_INNER_FOR_BODY75:%.*]], label [[OMP_INNER_FOR_END100:%.*]]
8382 // CHECK15:       omp.inner.for.body75:
8383 // CHECK15-NEXT:    [[TMP46:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
8384 // CHECK15-NEXT:    [[MUL76:%.*]] = mul nsw i32 [[TMP46]], 1
8385 // CHECK15-NEXT:    [[SUB77:%.*]] = sub nsw i32 122, [[MUL76]]
8386 // CHECK15-NEXT:    [[CONV78:%.*]] = trunc i32 [[SUB77]] to i8
8387 // CHECK15-NEXT:    store i8 [[CONV78]], i8* [[IT72]], align 1, !llvm.access.group !16
8388 // CHECK15-NEXT:    [[TMP47:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !16
8389 // CHECK15-NEXT:    [[ADD79:%.*]] = add nsw i32 [[TMP47]], 1
8390 // CHECK15-NEXT:    store i32 [[ADD79]], i32* [[A]], align 4, !llvm.access.group !16
8391 // CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[B]], i32 0, i32 2
8392 // CHECK15-NEXT:    [[TMP48:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !16
8393 // CHECK15-NEXT:    [[CONV80:%.*]] = fpext float [[TMP48]] to double
8394 // CHECK15-NEXT:    [[ADD81:%.*]] = fadd double [[CONV80]], 1.000000e+00
8395 // CHECK15-NEXT:    [[CONV82:%.*]] = fptrunc double [[ADD81]] to float
8396 // CHECK15-NEXT:    store float [[CONV82]], float* [[ARRAYIDX]], align 4, !llvm.access.group !16
8397 // CHECK15-NEXT:    [[ARRAYIDX83:%.*]] = getelementptr inbounds float, float* [[VLA]], i32 3
8398 // CHECK15-NEXT:    [[TMP49:%.*]] = load float, float* [[ARRAYIDX83]], align 4, !llvm.access.group !16
8399 // CHECK15-NEXT:    [[CONV84:%.*]] = fpext float [[TMP49]] to double
8400 // CHECK15-NEXT:    [[ADD85:%.*]] = fadd double [[CONV84]], 1.000000e+00
8401 // CHECK15-NEXT:    [[CONV86:%.*]] = fptrunc double [[ADD85]] to float
8402 // CHECK15-NEXT:    store float [[CONV86]], float* [[ARRAYIDX83]], align 4, !llvm.access.group !16
8403 // CHECK15-NEXT:    [[ARRAYIDX87:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[C]], i32 0, i32 1
8404 // CHECK15-NEXT:    [[ARRAYIDX88:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX87]], i32 0, i32 2
8405 // CHECK15-NEXT:    [[TMP50:%.*]] = load double, double* [[ARRAYIDX88]], align 8, !llvm.access.group !16
8406 // CHECK15-NEXT:    [[ADD89:%.*]] = fadd double [[TMP50]], 1.000000e+00
8407 // CHECK15-NEXT:    store double [[ADD89]], double* [[ARRAYIDX88]], align 8, !llvm.access.group !16
8408 // CHECK15-NEXT:    [[TMP51:%.*]] = mul nsw i32 1, [[TMP2]]
8409 // CHECK15-NEXT:    [[ARRAYIDX90:%.*]] = getelementptr inbounds double, double* [[VLA1]], i32 [[TMP51]]
8410 // CHECK15-NEXT:    [[ARRAYIDX91:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX90]], i32 3
8411 // CHECK15-NEXT:    [[TMP52:%.*]] = load double, double* [[ARRAYIDX91]], align 8, !llvm.access.group !16
8412 // CHECK15-NEXT:    [[ADD92:%.*]] = fadd double [[TMP52]], 1.000000e+00
8413 // CHECK15-NEXT:    store double [[ADD92]], double* [[ARRAYIDX91]], align 8, !llvm.access.group !16
8414 // CHECK15-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 0
8415 // CHECK15-NEXT:    [[TMP53:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !16
8416 // CHECK15-NEXT:    [[ADD93:%.*]] = add nsw i64 [[TMP53]], 1
8417 // CHECK15-NEXT:    store i64 [[ADD93]], i64* [[X]], align 4, !llvm.access.group !16
8418 // CHECK15-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[D]], i32 0, i32 1
8419 // CHECK15-NEXT:    [[TMP54:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !16
8420 // CHECK15-NEXT:    [[CONV94:%.*]] = sext i8 [[TMP54]] to i32
8421 // CHECK15-NEXT:    [[ADD95:%.*]] = add nsw i32 [[CONV94]], 1
8422 // CHECK15-NEXT:    [[CONV96:%.*]] = trunc i32 [[ADD95]] to i8
8423 // CHECK15-NEXT:    store i8 [[CONV96]], i8* [[Y]], align 4, !llvm.access.group !16
8424 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE97:%.*]]
8425 // CHECK15:       omp.body.continue97:
8426 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC98:%.*]]
8427 // CHECK15:       omp.inner.for.inc98:
8428 // CHECK15-NEXT:    [[TMP55:%.*]] = load i32, i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
8429 // CHECK15-NEXT:    [[ADD99:%.*]] = add nsw i32 [[TMP55]], 1
8430 // CHECK15-NEXT:    store i32 [[ADD99]], i32* [[DOTOMP_IV71]], align 4, !llvm.access.group !16
8431 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND73]], !llvm.loop [[LOOP17:![0-9]+]]
8432 // CHECK15:       omp.inner.for.end100:
8433 // CHECK15-NEXT:    store i8 96, i8* [[IT72]], align 1
8434 // CHECK15-NEXT:    [[TMP56:%.*]] = load i32, i32* [[A]], align 4
8435 // CHECK15-NEXT:    [[TMP57:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
8436 // CHECK15-NEXT:    call void @llvm.stackrestore(i8* [[TMP57]])
8437 // CHECK15-NEXT:    ret i32 [[TMP56]]
8438 //
8439 //
8440 // CHECK15-LABEL: define {{[^@]+}}@_Z3bari
8441 // CHECK15-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
8442 // CHECK15-NEXT:  entry:
8443 // CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8444 // CHECK15-NEXT:    [[A:%.*]] = alloca i32, align 4
8445 // CHECK15-NEXT:    [[S:%.*]] = alloca [[STRUCT_S1:%.*]], align 4
8446 // CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8447 // CHECK15-NEXT:    store i32 0, i32* [[A]], align 4
8448 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8449 // CHECK15-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[TMP0]])
8450 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A]], align 4
8451 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP1]], [[CALL]]
8452 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[A]], align 4
8453 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
8454 // CHECK15-NEXT:    [[CALL1:%.*]] = call noundef i32 @_ZN2S12r1Ei(%struct.S1* noundef nonnull align 4 dereferenceable(8) [[S]], i32 noundef [[TMP2]])
8455 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A]], align 4
8456 // CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP3]], [[CALL1]]
8457 // CHECK15-NEXT:    store i32 [[ADD2]], i32* [[A]], align 4
8458 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8459 // CHECK15-NEXT:    [[CALL3:%.*]] = call noundef i32 @_ZL7fstatici(i32 noundef [[TMP4]])
8460 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A]], align 4
8461 // CHECK15-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP5]], [[CALL3]]
8462 // CHECK15-NEXT:    store i32 [[ADD4]], i32* [[A]], align 4
8463 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
8464 // CHECK15-NEXT:    [[CALL5:%.*]] = call noundef i32 @_Z9ftemplateIiET_i(i32 noundef [[TMP6]])
8465 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[A]], align 4
8466 // CHECK15-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP7]], [[CALL5]]
8467 // CHECK15-NEXT:    store i32 [[ADD6]], i32* [[A]], align 4
8468 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
8469 // CHECK15-NEXT:    ret i32 [[TMP8]]
8470 //
8471 //
8472 // CHECK15-LABEL: define {{[^@]+}}@_ZN2S12r1Ei
8473 // CHECK15-SAME: (%struct.S1* noundef nonnull align 4 dereferenceable(8) [[THIS:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] comdat align 2 {
8474 // CHECK15-NEXT:  entry:
8475 // CHECK15-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
8476 // CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8477 // CHECK15-NEXT:    [[B:%.*]] = alloca i32, align 4
8478 // CHECK15-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 4
8479 // CHECK15-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i32, align 4
8480 // CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
8481 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i64, align 4
8482 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
8483 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
8484 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
8485 // CHECK15-NEXT:    [[IT:%.*]] = alloca i64, align 8
8486 // CHECK15-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
8487 // CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8488 // CHECK15-NEXT:    [[THIS1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
8489 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
8490 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP0]], 1
8491 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[B]], align 4
8492 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
8493 // CHECK15-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
8494 // CHECK15-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
8495 // CHECK15-NEXT:    [[TMP3:%.*]] = mul nuw i32 2, [[TMP1]]
8496 // CHECK15-NEXT:    [[VLA:%.*]] = alloca i16, i32 [[TMP3]], align 2
8497 // CHECK15-NEXT:    store i32 [[TMP1]], i32* [[__VLA_EXPR0]], align 4
8498 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[N_ADDR]], align 4
8499 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 60
8500 // CHECK15-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[CMP]] to i8
8501 // CHECK15-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
8502 // CHECK15-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
8503 // CHECK15-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
8504 // CHECK15-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
8505 // CHECK15-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
8506 // CHECK15-NEXT:    [[TMP6:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
8507 // CHECK15-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP6]] to i1
8508 // CHECK15-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8509 // CHECK15:       omp_if.then:
8510 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8511 // CHECK15:       omp.inner.for.cond:
8512 // CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
8513 // CHECK15-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !19
8514 // CHECK15-NEXT:    [[CMP2:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
8515 // CHECK15-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8516 // CHECK15:       omp.inner.for.body:
8517 // CHECK15-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
8518 // CHECK15-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
8519 // CHECK15-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
8520 // CHECK15-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !19
8521 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[B]], align 4, !llvm.access.group !19
8522 // CHECK15-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP10]] to double
8523 // CHECK15-NEXT:    [[ADD3:%.*]] = fadd double [[CONV]], 1.500000e+00
8524 // CHECK15-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[THIS1]], i32 0, i32 0
8525 // CHECK15-NEXT:    store double [[ADD3]], double* [[A]], align 4, !nontemporal !20, !llvm.access.group !19
8526 // CHECK15-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8527 // CHECK15-NEXT:    [[TMP11:%.*]] = load double, double* [[A4]], align 4, !nontemporal !20, !llvm.access.group !19
8528 // CHECK15-NEXT:    [[INC:%.*]] = fadd double [[TMP11]], 1.000000e+00
8529 // CHECK15-NEXT:    store double [[INC]], double* [[A4]], align 4, !nontemporal !20, !llvm.access.group !19
8530 // CHECK15-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
8531 // CHECK15-NEXT:    [[TMP12:%.*]] = mul nsw i32 1, [[TMP1]]
8532 // CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP12]]
8533 // CHECK15-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
8534 // CHECK15-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2, !llvm.access.group !19
8535 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8536 // CHECK15:       omp.body.continue:
8537 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8538 // CHECK15:       omp.inner.for.inc:
8539 // CHECK15-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
8540 // CHECK15-NEXT:    [[ADD7:%.*]] = add i64 [[TMP13]], 1
8541 // CHECK15-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !19
8542 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
8543 // CHECK15:       omp.inner.for.end:
8544 // CHECK15-NEXT:    br label [[OMP_IF_END:%.*]]
8545 // CHECK15:       omp_if.else:
8546 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
8547 // CHECK15:       omp.inner.for.cond8:
8548 // CHECK15-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8549 // CHECK15-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8550 // CHECK15-NEXT:    [[CMP9:%.*]] = icmp ule i64 [[TMP14]], [[TMP15]]
8551 // CHECK15-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END24:%.*]]
8552 // CHECK15:       omp.inner.for.body10:
8553 // CHECK15-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8554 // CHECK15-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP16]], 400
8555 // CHECK15-NEXT:    [[SUB12:%.*]] = sub i64 2000, [[MUL11]]
8556 // CHECK15-NEXT:    store i64 [[SUB12]], i64* [[IT]], align 8
8557 // CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[B]], align 4
8558 // CHECK15-NEXT:    [[CONV13:%.*]] = sitofp i32 [[TMP17]] to double
8559 // CHECK15-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.500000e+00
8560 // CHECK15-NEXT:    [[A15:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8561 // CHECK15-NEXT:    store double [[ADD14]], double* [[A15]], align 4
8562 // CHECK15-NEXT:    [[A16:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[THIS1]], i32 0, i32 0
8563 // CHECK15-NEXT:    [[TMP18:%.*]] = load double, double* [[A16]], align 4
8564 // CHECK15-NEXT:    [[INC17:%.*]] = fadd double [[TMP18]], 1.000000e+00
8565 // CHECK15-NEXT:    store double [[INC17]], double* [[A16]], align 4
8566 // CHECK15-NEXT:    [[CONV18:%.*]] = fptosi double [[INC17]] to i16
8567 // CHECK15-NEXT:    [[TMP19:%.*]] = mul nsw i32 1, [[TMP1]]
8568 // CHECK15-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP19]]
8569 // CHECK15-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX19]], i32 1
8570 // CHECK15-NEXT:    store i16 [[CONV18]], i16* [[ARRAYIDX20]], align 2
8571 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE21:%.*]]
8572 // CHECK15:       omp.body.continue21:
8573 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC22:%.*]]
8574 // CHECK15:       omp.inner.for.inc22:
8575 // CHECK15-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8576 // CHECK15-NEXT:    [[ADD23:%.*]] = add i64 [[TMP20]], 1
8577 // CHECK15-NEXT:    store i64 [[ADD23]], i64* [[DOTOMP_IV]], align 8
8578 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP23:![0-9]+]]
8579 // CHECK15:       omp.inner.for.end24:
8580 // CHECK15-NEXT:    br label [[OMP_IF_END]]
8581 // CHECK15:       omp_if.end:
8582 // CHECK15-NEXT:    store i64 400, i64* [[IT]], align 8
8583 // CHECK15-NEXT:    [[TMP21:%.*]] = mul nsw i32 1, [[TMP1]]
8584 // CHECK15-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds i16, i16* [[VLA]], i32 [[TMP21]]
8585 // CHECK15-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX25]], i32 1
8586 // CHECK15-NEXT:    [[TMP22:%.*]] = load i16, i16* [[ARRAYIDX26]], align 2
8587 // CHECK15-NEXT:    [[CONV27:%.*]] = sext i16 [[TMP22]] to i32
8588 // CHECK15-NEXT:    [[TMP23:%.*]] = load i32, i32* [[B]], align 4
8589 // CHECK15-NEXT:    [[ADD28:%.*]] = add nsw i32 [[CONV27]], [[TMP23]]
8590 // CHECK15-NEXT:    [[TMP24:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
8591 // CHECK15-NEXT:    call void @llvm.stackrestore(i8* [[TMP24]])
8592 // CHECK15-NEXT:    ret i32 [[ADD28]]
8593 //
8594 //
8595 // CHECK15-LABEL: define {{[^@]+}}@_ZL7fstatici
8596 // CHECK15-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] {
8597 // CHECK15-NEXT:  entry:
8598 // CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8599 // CHECK15-NEXT:    [[A:%.*]] = alloca i32, align 4
8600 // CHECK15-NEXT:    [[AA:%.*]] = alloca i16, align 2
8601 // CHECK15-NEXT:    [[AAA:%.*]] = alloca i8, align 1
8602 // CHECK15-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8603 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8604 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8605 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8606 // CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8607 // CHECK15-NEXT:    store i32 0, i32* [[A]], align 4
8608 // CHECK15-NEXT:    store i16 0, i16* [[AA]], align 2
8609 // CHECK15-NEXT:    store i8 0, i8* [[AAA]], align 1
8610 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8611 // CHECK15-NEXT:    store i32 429496720, i32* [[DOTOMP_UB]], align 4
8612 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A]], align 4
8613 // CHECK15-NEXT:    ret i32 [[TMP0]]
8614 //
8615 //
8616 // CHECK15-LABEL: define {{[^@]+}}@_Z9ftemplateIiET_i
8617 // CHECK15-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0]] comdat {
8618 // CHECK15-NEXT:  entry:
8619 // CHECK15-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
8620 // CHECK15-NEXT:    [[A:%.*]] = alloca i32, align 4
8621 // CHECK15-NEXT:    [[AA:%.*]] = alloca i16, align 2
8622 // CHECK15-NEXT:    [[B:%.*]] = alloca [10 x i32], align 4
8623 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i64, align 4
8624 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
8625 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
8626 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
8627 // CHECK15-NEXT:    [[I:%.*]] = alloca i64, align 8
8628 // CHECK15-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
8629 // CHECK15-NEXT:    store i32 0, i32* [[A]], align 4
8630 // CHECK15-NEXT:    store i16 0, i16* [[AA]], align 2
8631 // CHECK15-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
8632 // CHECK15-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
8633 // CHECK15-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
8634 // CHECK15-NEXT:    store i64 [[TMP0]], i64* [[DOTOMP_IV]], align 8
8635 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8636 // CHECK15:       omp.inner.for.cond:
8637 // CHECK15-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !25
8638 // CHECK15-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !25
8639 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i64 [[TMP1]], [[TMP2]]
8640 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8641 // CHECK15:       omp.inner.for.body:
8642 // CHECK15-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !25
8643 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP3]], 3
8644 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
8645 // CHECK15-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !25
8646 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A]], align 4, !llvm.access.group !25
8647 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
8648 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[A]], align 4, !llvm.access.group !25
8649 // CHECK15-NEXT:    [[TMP5:%.*]] = load i16, i16* [[AA]], align 2, !llvm.access.group !25
8650 // CHECK15-NEXT:    [[CONV:%.*]] = sext i16 [[TMP5]] to i32
8651 // CHECK15-NEXT:    [[ADD2:%.*]] = add nsw i32 [[CONV]], 1
8652 // CHECK15-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD2]] to i16
8653 // CHECK15-NEXT:    store i16 [[CONV3]], i16* [[AA]], align 2, !llvm.access.group !25
8654 // CHECK15-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[B]], i32 0, i32 2
8655 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
8656 // CHECK15-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP6]], 1
8657 // CHECK15-NEXT:    store i32 [[ADD4]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
8658 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8659 // CHECK15:       omp.body.continue:
8660 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8661 // CHECK15:       omp.inner.for.inc:
8662 // CHECK15-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !25
8663 // CHECK15-NEXT:    [[ADD5:%.*]] = add nsw i64 [[TMP7]], 1
8664 // CHECK15-NEXT:    store i64 [[ADD5]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !25
8665 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
8666 // CHECK15:       omp.inner.for.end:
8667 // CHECK15-NEXT:    store i64 11, i64* [[I]], align 8
8668 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A]], align 4
8669 // CHECK15-NEXT:    ret i32 [[TMP8]]
8670 //
8671 //
8672 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
8673 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] {
8674 // CHECK17-NEXT:  entry:
8675 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
8676 // CHECK17-NEXT:    ret void
8677 //
8678 //
8679 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
8680 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
8681 // CHECK17-NEXT:  entry:
8682 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8683 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8684 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8685 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8686 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8687 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8688 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8689 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8690 // CHECK17-NEXT:    [[I:%.*]] = alloca i32, align 4
8691 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8692 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8693 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8694 // CHECK17-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
8695 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8696 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8697 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8698 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8699 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8700 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8701 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
8702 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8703 // CHECK17:       cond.true:
8704 // CHECK17-NEXT:    br label [[COND_END:%.*]]
8705 // CHECK17:       cond.false:
8706 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8707 // CHECK17-NEXT:    br label [[COND_END]]
8708 // CHECK17:       cond.end:
8709 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8710 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8711 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8712 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8713 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8714 // CHECK17:       omp.inner.for.cond:
8715 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
8716 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
8717 // CHECK17-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8718 // CHECK17-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8719 // CHECK17:       omp.inner.for.body:
8720 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
8721 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
8722 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
8723 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
8724 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8725 // CHECK17:       omp.body.continue:
8726 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8727 // CHECK17:       omp.inner.for.inc:
8728 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
8729 // CHECK17-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
8730 // CHECK17-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
8731 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
8732 // CHECK17:       omp.inner.for.end:
8733 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8734 // CHECK17:       omp.loop.exit:
8735 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8736 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8737 // CHECK17-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
8738 // CHECK17-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8739 // CHECK17:       .omp.final.then:
8740 // CHECK17-NEXT:    store i32 33, i32* [[I]], align 4
8741 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8742 // CHECK17:       .omp.final.done:
8743 // CHECK17-NEXT:    ret void
8744 //
8745 //
8746 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
8747 // CHECK17-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
8748 // CHECK17-NEXT:  entry:
8749 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8750 // CHECK17-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
8751 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8752 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8753 // CHECK17-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
8754 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8755 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8756 // CHECK17-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
8757 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8758 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8759 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
8760 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8761 // CHECK17-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
8762 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8763 // CHECK17-NEXT:    store i16 [[TMP0]], i16* [[CONV3]], align 2
8764 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8765 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
8766 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
8767 // CHECK17-NEXT:    store i32 [[TMP2]], i32* [[CONV4]], align 4
8768 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
8769 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV2]], align 4
8770 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8771 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
8772 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[A_CASTED]], align 8
8773 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]])
8774 // CHECK17-NEXT:    ret void
8775 //
8776 //
8777 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
8778 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
8779 // CHECK17-NEXT:  entry:
8780 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8781 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8782 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8783 // CHECK17-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
8784 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8785 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
8786 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i64, align 8
8787 // CHECK17-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
8788 // CHECK17-NEXT:    [[DOTLINEAR_START3:%.*]] = alloca i32, align 4
8789 // CHECK17-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
8790 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
8791 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
8792 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
8793 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8794 // CHECK17-NEXT:    [[IT:%.*]] = alloca i64, align 8
8795 // CHECK17-NEXT:    [[LIN4:%.*]] = alloca i32, align 4
8796 // CHECK17-NEXT:    [[A5:%.*]] = alloca i32, align 4
8797 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8798 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8799 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8800 // CHECK17-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
8801 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8802 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8803 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
8804 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8805 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV1]], align 4
8806 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
8807 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV2]], align 4
8808 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START3]], align 4
8809 // CHECK17-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv() #[[ATTR5:[0-9]+]]
8810 // CHECK17-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
8811 // CHECK17-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
8812 // CHECK17-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
8813 // CHECK17-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
8814 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8815 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8816 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8817 // CHECK17-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
8818 // CHECK17-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
8819 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8820 // CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
8821 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8822 // CHECK17:       cond.true:
8823 // CHECK17-NEXT:    br label [[COND_END:%.*]]
8824 // CHECK17:       cond.false:
8825 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8826 // CHECK17-NEXT:    br label [[COND_END]]
8827 // CHECK17:       cond.end:
8828 // CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8829 // CHECK17-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
8830 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
8831 // CHECK17-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
8832 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8833 // CHECK17:       omp.inner.for.cond:
8834 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8835 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !17
8836 // CHECK17-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
8837 // CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8838 // CHECK17:       omp.inner.for.body:
8839 // CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8840 // CHECK17-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
8841 // CHECK17-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
8842 // CHECK17-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !17
8843 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !17
8844 // CHECK17-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP10]] to i64
8845 // CHECK17-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8846 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !17
8847 // CHECK17-NEXT:    [[MUL8:%.*]] = mul i64 [[TMP11]], [[TMP12]]
8848 // CHECK17-NEXT:    [[ADD:%.*]] = add i64 [[CONV7]], [[MUL8]]
8849 // CHECK17-NEXT:    [[CONV9:%.*]] = trunc i64 [[ADD]] to i32
8850 // CHECK17-NEXT:    store i32 [[CONV9]], i32* [[LIN4]], align 4, !llvm.access.group !17
8851 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START3]], align 4, !llvm.access.group !17
8852 // CHECK17-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
8853 // CHECK17-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8854 // CHECK17-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !17
8855 // CHECK17-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP14]], [[TMP15]]
8856 // CHECK17-NEXT:    [[ADD12:%.*]] = add i64 [[CONV10]], [[MUL11]]
8857 // CHECK17-NEXT:    [[CONV13:%.*]] = trunc i64 [[ADD12]] to i32
8858 // CHECK17-NEXT:    store i32 [[CONV13]], i32* [[A5]], align 4, !llvm.access.group !17
8859 // CHECK17-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !17
8860 // CHECK17-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP16]] to i32
8861 // CHECK17-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
8862 // CHECK17-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
8863 // CHECK17-NEXT:    store i16 [[CONV16]], i16* [[CONV]], align 2, !llvm.access.group !17
8864 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8865 // CHECK17:       omp.body.continue:
8866 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8867 // CHECK17:       omp.inner.for.inc:
8868 // CHECK17-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8869 // CHECK17-NEXT:    [[ADD17:%.*]] = add i64 [[TMP17]], 1
8870 // CHECK17-NEXT:    store i64 [[ADD17]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
8871 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
8872 // CHECK17:       omp.inner.for.end:
8873 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8874 // CHECK17:       omp.loop.exit:
8875 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8876 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8877 // CHECK17-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
8878 // CHECK17-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8879 // CHECK17:       .omp.final.then:
8880 // CHECK17-NEXT:    store i64 400, i64* [[IT]], align 8
8881 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8882 // CHECK17:       .omp.final.done:
8883 // CHECK17-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8884 // CHECK17-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
8885 // CHECK17-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
8886 // CHECK17:       .omp.linear.pu:
8887 // CHECK17-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN4]], align 4
8888 // CHECK17-NEXT:    store i32 [[TMP22]], i32* [[CONV1]], align 4
8889 // CHECK17-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A5]], align 4
8890 // CHECK17-NEXT:    store i32 [[TMP23]], i32* [[CONV2]], align 4
8891 // CHECK17-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
8892 // CHECK17:       .omp.linear.pu.done:
8893 // CHECK17-NEXT:    ret void
8894 //
8895 //
8896 // CHECK17-LABEL: define {{[^@]+}}@_Z7get_valv
8897 // CHECK17-SAME: () #[[ATTR3:[0-9]+]] {
8898 // CHECK17-NEXT:  entry:
8899 // CHECK17-NEXT:    ret i64 0
8900 //
8901 //
8902 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
8903 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
8904 // CHECK17-NEXT:  entry:
8905 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8906 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8907 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
8908 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
8909 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8910 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8911 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8912 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8913 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
8914 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
8915 // CHECK17-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
8916 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
8917 // CHECK17-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
8918 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
8919 // CHECK17-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
8920 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
8921 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
8922 // CHECK17-NEXT:    ret void
8923 //
8924 //
8925 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
8926 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR1]] {
8927 // CHECK17-NEXT:  entry:
8928 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8929 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8930 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
8931 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
8932 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8933 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i16, align 2
8934 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8935 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8936 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8937 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8938 // CHECK17-NEXT:    [[IT:%.*]] = alloca i16, align 2
8939 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8940 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8941 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
8942 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
8943 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
8944 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
8945 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8946 // CHECK17-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
8947 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8948 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8949 // CHECK17-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8950 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8951 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8952 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8953 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
8954 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8955 // CHECK17:       cond.true:
8956 // CHECK17-NEXT:    br label [[COND_END:%.*]]
8957 // CHECK17:       cond.false:
8958 // CHECK17-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8959 // CHECK17-NEXT:    br label [[COND_END]]
8960 // CHECK17:       cond.end:
8961 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8962 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8963 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8964 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8965 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8966 // CHECK17:       omp.inner.for.cond:
8967 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
8968 // CHECK17-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
8969 // CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8970 // CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8971 // CHECK17:       omp.inner.for.body:
8972 // CHECK17-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
8973 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
8974 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
8975 // CHECK17-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i16
8976 // CHECK17-NEXT:    store i16 [[CONV3]], i16* [[IT]], align 2, !llvm.access.group !20
8977 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !20
8978 // CHECK17-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
8979 // CHECK17-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !20
8980 // CHECK17-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !20
8981 // CHECK17-NEXT:    [[CONV5:%.*]] = sext i16 [[TMP9]] to i32
8982 // CHECK17-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
8983 // CHECK17-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i16
8984 // CHECK17-NEXT:    store i16 [[CONV7]], i16* [[CONV1]], align 2, !llvm.access.group !20
8985 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8986 // CHECK17:       omp.body.continue:
8987 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8988 // CHECK17:       omp.inner.for.inc:
8989 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
8990 // CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 1
8991 // CHECK17-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
8992 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
8993 // CHECK17:       omp.inner.for.end:
8994 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8995 // CHECK17:       omp.loop.exit:
8996 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8997 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8998 // CHECK17-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8999 // CHECK17-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9000 // CHECK17:       .omp.final.then:
9001 // CHECK17-NEXT:    store i16 22, i16* [[IT]], align 2
9002 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9003 // CHECK17:       .omp.final.done:
9004 // CHECK17-NEXT:    ret void
9005 //
9006 //
9007 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
9008 // CHECK17-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
9009 // CHECK17-NEXT:  entry:
9010 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9011 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
9012 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9013 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
9014 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
9015 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9016 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
9017 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
9018 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
9019 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9020 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9021 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9022 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9023 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
9024 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9025 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
9026 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
9027 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9028 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
9029 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
9030 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
9031 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9032 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9033 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
9034 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9035 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
9036 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
9037 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9038 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
9039 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
9040 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
9041 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9042 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
9043 // CHECK17-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9044 // CHECK17-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
9045 // CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
9046 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
9047 // CHECK17-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9048 // CHECK17-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
9049 // CHECK17-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9050 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
9051 // CHECK17-NEXT:    ret void
9052 //
9053 //
9054 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
9055 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
9056 // CHECK17-NEXT:  entry:
9057 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9058 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9059 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9060 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
9061 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9062 // CHECK17-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
9063 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
9064 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9065 // CHECK17-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
9066 // CHECK17-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
9067 // CHECK17-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
9068 // CHECK17-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9069 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9070 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i8, align 1
9071 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9072 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9073 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9074 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9075 // CHECK17-NEXT:    [[IT:%.*]] = alloca i8, align 1
9076 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9077 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9078 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9079 // CHECK17-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
9080 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9081 // CHECK17-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
9082 // CHECK17-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
9083 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9084 // CHECK17-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
9085 // CHECK17-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
9086 // CHECK17-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
9087 // CHECK17-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9088 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9089 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
9090 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9091 // CHECK17-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
9092 // CHECK17-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
9093 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9094 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
9095 // CHECK17-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
9096 // CHECK17-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
9097 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9098 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9099 // CHECK17-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
9100 // CHECK17-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9101 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9102 // CHECK17-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
9103 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9104 // CHECK17-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
9105 // CHECK17-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
9106 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
9107 // CHECK17:       omp.dispatch.cond:
9108 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9109 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
9110 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9111 // CHECK17:       cond.true:
9112 // CHECK17-NEXT:    br label [[COND_END:%.*]]
9113 // CHECK17:       cond.false:
9114 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9115 // CHECK17-NEXT:    br label [[COND_END]]
9116 // CHECK17:       cond.end:
9117 // CHECK17-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
9118 // CHECK17-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9119 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9120 // CHECK17-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
9121 // CHECK17-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9122 // CHECK17-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9123 // CHECK17-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
9124 // CHECK17-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9125 // CHECK17:       omp.dispatch.body:
9126 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9127 // CHECK17:       omp.inner.for.cond:
9128 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
9129 // CHECK17-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
9130 // CHECK17-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
9131 // CHECK17-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9132 // CHECK17:       omp.inner.for.body:
9133 // CHECK17-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
9134 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
9135 // CHECK17-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
9136 // CHECK17-NEXT:    [[CONV8:%.*]] = trunc i32 [[SUB]] to i8
9137 // CHECK17-NEXT:    store i8 [[CONV8]], i8* [[IT]], align 1, !llvm.access.group !23
9138 // CHECK17-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !23
9139 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
9140 // CHECK17-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4, !llvm.access.group !23
9141 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
9142 // CHECK17-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
9143 // CHECK17-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
9144 // CHECK17-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
9145 // CHECK17-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
9146 // CHECK17-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
9147 // CHECK17-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
9148 // CHECK17-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !23
9149 // CHECK17-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
9150 // CHECK17-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
9151 // CHECK17-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
9152 // CHECK17-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !23
9153 // CHECK17-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
9154 // CHECK17-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
9155 // CHECK17-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !23
9156 // CHECK17-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
9157 // CHECK17-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !23
9158 // CHECK17-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
9159 // CHECK17-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
9160 // CHECK17-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
9161 // CHECK17-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !23
9162 // CHECK17-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
9163 // CHECK17-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !23
9164 // CHECK17-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
9165 // CHECK17-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !23
9166 // CHECK17-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
9167 // CHECK17-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !23
9168 // CHECK17-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
9169 // CHECK17-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !23
9170 // CHECK17-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
9171 // CHECK17-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
9172 // CHECK17-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
9173 // CHECK17-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !23
9174 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9175 // CHECK17:       omp.body.continue:
9176 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9177 // CHECK17:       omp.inner.for.inc:
9178 // CHECK17-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
9179 // CHECK17-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
9180 // CHECK17-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
9181 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
9182 // CHECK17:       omp.inner.for.end:
9183 // CHECK17-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
9184 // CHECK17:       omp.dispatch.inc:
9185 // CHECK17-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9186 // CHECK17-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9187 // CHECK17-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
9188 // CHECK17-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
9189 // CHECK17-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9190 // CHECK17-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9191 // CHECK17-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
9192 // CHECK17-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
9193 // CHECK17-NEXT:    br label [[OMP_DISPATCH_COND]]
9194 // CHECK17:       omp.dispatch.end:
9195 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
9196 // CHECK17-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9197 // CHECK17-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
9198 // CHECK17-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9199 // CHECK17:       .omp.final.then:
9200 // CHECK17-NEXT:    store i8 96, i8* [[IT]], align 1
9201 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9202 // CHECK17:       .omp.final.done:
9203 // CHECK17-NEXT:    ret void
9204 //
9205 //
9206 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
9207 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9208 // CHECK17-NEXT:  entry:
9209 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9210 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9211 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
9212 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9213 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9214 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9215 // CHECK17-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
9216 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9217 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9218 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
9219 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9220 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9221 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9222 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
9223 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9224 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9225 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9226 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
9227 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
9228 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
9229 // CHECK17-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9230 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
9231 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9232 // CHECK17-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
9233 // CHECK17-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
9234 // CHECK17-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
9235 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
9236 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
9237 // CHECK17-NEXT:    ret void
9238 //
9239 //
9240 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..4
9241 // 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:%.*]]) #[[ATTR1]] {
9242 // CHECK17-NEXT:  entry:
9243 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9244 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9245 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9246 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9247 // CHECK17-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
9248 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9249 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9250 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9251 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9252 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9253 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9254 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9255 // CHECK17-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
9256 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9257 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9258 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9259 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
9260 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9261 // CHECK17-NEXT:    ret void
9262 //
9263 //
9264 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
9265 // CHECK17-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
9266 // CHECK17-NEXT:  entry:
9267 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
9268 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
9269 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9270 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9271 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
9272 // CHECK17-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
9273 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
9274 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
9275 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9276 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9277 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
9278 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
9279 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
9280 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9281 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9282 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
9283 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV]], align 4
9284 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[B_CASTED]] to i32*
9285 // CHECK17-NEXT:    store i32 [[TMP4]], i32* [[CONV3]], align 4
9286 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[B_CASTED]], align 8
9287 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i64 [[TMP5]], i64 [[TMP1]], i64 [[TMP2]], i16* [[TMP3]])
9288 // CHECK17-NEXT:    ret void
9289 //
9290 //
9291 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..5
9292 // 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:%.*]]) #[[ATTR1]] {
9293 // CHECK17-NEXT:  entry:
9294 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9295 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9296 // CHECK17-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
9297 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
9298 // CHECK17-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
9299 // CHECK17-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
9300 // CHECK17-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
9301 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
9302 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i64, align 8
9303 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
9304 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
9305 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
9306 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9307 // CHECK17-NEXT:    [[IT:%.*]] = alloca i64, align 8
9308 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9309 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9310 // CHECK17-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
9311 // CHECK17-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
9312 // CHECK17-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
9313 // CHECK17-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
9314 // CHECK17-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
9315 // CHECK17-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
9316 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
9317 // CHECK17-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
9318 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
9319 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
9320 // CHECK17-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
9321 // CHECK17-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
9322 // CHECK17-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
9323 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9324 // CHECK17-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9325 // CHECK17-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
9326 // CHECK17-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
9327 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9328 // CHECK17-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP6]], 3
9329 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9330 // CHECK17:       cond.true:
9331 // CHECK17-NEXT:    br label [[COND_END:%.*]]
9332 // CHECK17:       cond.false:
9333 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9334 // CHECK17-NEXT:    br label [[COND_END]]
9335 // CHECK17:       cond.end:
9336 // CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
9337 // CHECK17-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
9338 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
9339 // CHECK17-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
9340 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9341 // CHECK17:       omp.inner.for.cond:
9342 // CHECK17-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
9343 // CHECK17-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !26
9344 // CHECK17-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP9]], [[TMP10]]
9345 // CHECK17-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9346 // CHECK17:       omp.inner.for.body:
9347 // CHECK17-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
9348 // CHECK17-NEXT:    [[MUL:%.*]] = mul i64 [[TMP11]], 400
9349 // CHECK17-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
9350 // CHECK17-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !26
9351 // CHECK17-NEXT:    [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !26
9352 // CHECK17-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP12]] to double
9353 // CHECK17-NEXT:    [[ADD:%.*]] = fadd double [[CONV4]], 1.500000e+00
9354 // CHECK17-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
9355 // CHECK17-NEXT:    store double [[ADD]], double* [[A]], align 8, !llvm.access.group !26
9356 // CHECK17-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
9357 // CHECK17-NEXT:    [[TMP13:%.*]] = load double, double* [[A5]], align 8, !llvm.access.group !26
9358 // CHECK17-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
9359 // CHECK17-NEXT:    store double [[INC]], double* [[A5]], align 8, !llvm.access.group !26
9360 // CHECK17-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
9361 // CHECK17-NEXT:    [[TMP14:%.*]] = mul nsw i64 1, [[TMP2]]
9362 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP14]]
9363 // CHECK17-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
9364 // CHECK17-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2, !llvm.access.group !26
9365 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9366 // CHECK17:       omp.body.continue:
9367 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9368 // CHECK17:       omp.inner.for.inc:
9369 // CHECK17-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
9370 // CHECK17-NEXT:    [[ADD8:%.*]] = add i64 [[TMP15]], 1
9371 // CHECK17-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
9372 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
9373 // CHECK17:       omp.inner.for.end:
9374 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9375 // CHECK17:       omp.loop.exit:
9376 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
9377 // CHECK17-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9378 // CHECK17-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
9379 // CHECK17-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9380 // CHECK17:       .omp.final.then:
9381 // CHECK17-NEXT:    store i64 400, i64* [[IT]], align 8
9382 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9383 // CHECK17:       .omp.final.done:
9384 // CHECK17-NEXT:    ret void
9385 //
9386 //
9387 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
9388 // CHECK17-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
9389 // CHECK17-NEXT:  entry:
9390 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9391 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9392 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9393 // CHECK17-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
9394 // CHECK17-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
9395 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9396 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9397 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9398 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9399 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9400 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9401 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
9402 // CHECK17-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
9403 // CHECK17-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
9404 // CHECK17-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
9405 // CHECK17-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
9406 // CHECK17-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
9407 // CHECK17-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
9408 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
9409 // CHECK17-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
9410 // CHECK17-NEXT:    ret void
9411 //
9412 //
9413 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
9414 // 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:%.*]]) #[[ATTR1]] {
9415 // CHECK17-NEXT:  entry:
9416 // CHECK17-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9417 // CHECK17-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9418 // CHECK17-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
9419 // CHECK17-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
9420 // CHECK17-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
9421 // CHECK17-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
9422 // CHECK17-NEXT:    [[TMP:%.*]] = alloca i64, align 8
9423 // CHECK17-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
9424 // CHECK17-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
9425 // CHECK17-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
9426 // CHECK17-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9427 // CHECK17-NEXT:    [[I:%.*]] = alloca i64, align 8
9428 // CHECK17-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9429 // CHECK17-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9430 // CHECK17-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
9431 // CHECK17-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
9432 // CHECK17-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
9433 // CHECK17-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
9434 // CHECK17-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
9435 // CHECK17-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
9436 // CHECK17-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
9437 // CHECK17-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
9438 // CHECK17-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
9439 // CHECK17-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9440 // CHECK17-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9441 // CHECK17-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9442 // CHECK17-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
9443 // CHECK17-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9444 // CHECK17-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
9445 // CHECK17-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9446 // CHECK17:       cond.true:
9447 // CHECK17-NEXT:    br label [[COND_END:%.*]]
9448 // CHECK17:       cond.false:
9449 // CHECK17-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9450 // CHECK17-NEXT:    br label [[COND_END]]
9451 // CHECK17:       cond.end:
9452 // CHECK17-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9453 // CHECK17-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
9454 // CHECK17-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
9455 // CHECK17-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
9456 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9457 // CHECK17:       omp.inner.for.cond:
9458 // CHECK17-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
9459 // CHECK17-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !29
9460 // CHECK17-NEXT:    [[CMP2:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
9461 // CHECK17-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9462 // CHECK17:       omp.inner.for.body:
9463 // CHECK17-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
9464 // CHECK17-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
9465 // CHECK17-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
9466 // CHECK17-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !29
9467 // CHECK17-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !29
9468 // CHECK17-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
9469 // CHECK17-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4, !llvm.access.group !29
9470 // CHECK17-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !29
9471 // CHECK17-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
9472 // CHECK17-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
9473 // CHECK17-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
9474 // CHECK17-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2, !llvm.access.group !29
9475 // CHECK17-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
9476 // CHECK17-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !29
9477 // CHECK17-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
9478 // CHECK17-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !29
9479 // CHECK17-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9480 // CHECK17:       omp.body.continue:
9481 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9482 // CHECK17:       omp.inner.for.inc:
9483 // CHECK17-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
9484 // CHECK17-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP12]], 1
9485 // CHECK17-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !29
9486 // CHECK17-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
9487 // CHECK17:       omp.inner.for.end:
9488 // CHECK17-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9489 // CHECK17:       omp.loop.exit:
9490 // CHECK17-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9491 // CHECK17-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9492 // CHECK17-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9493 // CHECK17-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9494 // CHECK17:       .omp.final.then:
9495 // CHECK17-NEXT:    store i64 11, i64* [[I]], align 8
9496 // CHECK17-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9497 // CHECK17:       .omp.final.done:
9498 // CHECK17-NEXT:    ret void
9499 //
9500 //
9501 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
9502 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] {
9503 // CHECK19-NEXT:  entry:
9504 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
9505 // CHECK19-NEXT:    ret void
9506 //
9507 //
9508 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
9509 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
9510 // CHECK19-NEXT:  entry:
9511 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9512 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9513 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9514 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9515 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9516 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9517 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9518 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9519 // CHECK19-NEXT:    [[I:%.*]] = alloca i32, align 4
9520 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9521 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9522 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9523 // CHECK19-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
9524 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9525 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9526 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9527 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9528 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9529 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9530 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
9531 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9532 // CHECK19:       cond.true:
9533 // CHECK19-NEXT:    br label [[COND_END:%.*]]
9534 // CHECK19:       cond.false:
9535 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9536 // CHECK19-NEXT:    br label [[COND_END]]
9537 // CHECK19:       cond.end:
9538 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9539 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9540 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9541 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9542 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9543 // CHECK19:       omp.inner.for.cond:
9544 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9545 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
9546 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9547 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9548 // CHECK19:       omp.inner.for.body:
9549 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9550 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
9551 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
9552 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
9553 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9554 // CHECK19:       omp.body.continue:
9555 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9556 // CHECK19:       omp.inner.for.inc:
9557 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9558 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
9559 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9560 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
9561 // CHECK19:       omp.inner.for.end:
9562 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9563 // CHECK19:       omp.loop.exit:
9564 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9565 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9566 // CHECK19-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
9567 // CHECK19-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9568 // CHECK19:       .omp.final.then:
9569 // CHECK19-NEXT:    store i32 33, i32* [[I]], align 4
9570 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9571 // CHECK19:       .omp.final.done:
9572 // CHECK19-NEXT:    ret void
9573 //
9574 //
9575 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
9576 // CHECK19-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
9577 // CHECK19-NEXT:  entry:
9578 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9579 // CHECK19-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
9580 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9581 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9582 // CHECK19-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
9583 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9584 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9585 // CHECK19-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
9586 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9587 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9588 // CHECK19-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
9589 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9590 // CHECK19-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
9591 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9592 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
9593 // CHECK19-NEXT:    store i32 [[TMP2]], i32* [[LIN_CASTED]], align 4
9594 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
9595 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
9596 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[A_CASTED]], align 4
9597 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A_CASTED]], align 4
9598 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]])
9599 // CHECK19-NEXT:    ret void
9600 //
9601 //
9602 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
9603 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
9604 // CHECK19-NEXT:  entry:
9605 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9606 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9607 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9608 // CHECK19-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
9609 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9610 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
9611 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i64, align 4
9612 // CHECK19-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
9613 // CHECK19-NEXT:    [[DOTLINEAR_START1:%.*]] = alloca i32, align 4
9614 // CHECK19-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
9615 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
9616 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
9617 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
9618 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9619 // CHECK19-NEXT:    [[IT:%.*]] = alloca i64, align 8
9620 // CHECK19-NEXT:    [[LIN2:%.*]] = alloca i32, align 4
9621 // CHECK19-NEXT:    [[A3:%.*]] = alloca i32, align 4
9622 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9623 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9624 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9625 // CHECK19-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
9626 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9627 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9628 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
9629 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
9630 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
9631 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START1]], align 4
9632 // CHECK19-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv() #[[ATTR5:[0-9]+]]
9633 // CHECK19-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
9634 // CHECK19-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
9635 // CHECK19-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
9636 // CHECK19-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
9637 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9638 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9639 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9640 // CHECK19-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
9641 // CHECK19-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
9642 // CHECK19-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9643 // CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
9644 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9645 // CHECK19:       cond.true:
9646 // CHECK19-NEXT:    br label [[COND_END:%.*]]
9647 // CHECK19:       cond.false:
9648 // CHECK19-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
9649 // CHECK19-NEXT:    br label [[COND_END]]
9650 // CHECK19:       cond.end:
9651 // CHECK19-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9652 // CHECK19-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
9653 // CHECK19-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
9654 // CHECK19-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
9655 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9656 // CHECK19:       omp.inner.for.cond:
9657 // CHECK19-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9658 // CHECK19-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !18
9659 // CHECK19-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
9660 // CHECK19-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9661 // CHECK19:       omp.inner.for.body:
9662 // CHECK19-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9663 // CHECK19-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
9664 // CHECK19-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
9665 // CHECK19-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !18
9666 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !18
9667 // CHECK19-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
9668 // CHECK19-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9669 // CHECK19-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !18
9670 // CHECK19-NEXT:    [[MUL6:%.*]] = mul i64 [[TMP11]], [[TMP12]]
9671 // CHECK19-NEXT:    [[ADD:%.*]] = add i64 [[CONV5]], [[MUL6]]
9672 // CHECK19-NEXT:    [[CONV7:%.*]] = trunc i64 [[ADD]] to i32
9673 // CHECK19-NEXT:    store i32 [[CONV7]], i32* [[LIN2]], align 4, !llvm.access.group !18
9674 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START1]], align 4, !llvm.access.group !18
9675 // CHECK19-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP13]] to i64
9676 // CHECK19-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9677 // CHECK19-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !18
9678 // CHECK19-NEXT:    [[MUL9:%.*]] = mul i64 [[TMP14]], [[TMP15]]
9679 // CHECK19-NEXT:    [[ADD10:%.*]] = add i64 [[CONV8]], [[MUL9]]
9680 // CHECK19-NEXT:    [[CONV11:%.*]] = trunc i64 [[ADD10]] to i32
9681 // CHECK19-NEXT:    store i32 [[CONV11]], i32* [[A3]], align 4, !llvm.access.group !18
9682 // CHECK19-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !18
9683 // CHECK19-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP16]] to i32
9684 // CHECK19-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
9685 // CHECK19-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
9686 // CHECK19-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2, !llvm.access.group !18
9687 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9688 // CHECK19:       omp.body.continue:
9689 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9690 // CHECK19:       omp.inner.for.inc:
9691 // CHECK19-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9692 // CHECK19-NEXT:    [[ADD15:%.*]] = add i64 [[TMP17]], 1
9693 // CHECK19-NEXT:    store i64 [[ADD15]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
9694 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
9695 // CHECK19:       omp.inner.for.end:
9696 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9697 // CHECK19:       omp.loop.exit:
9698 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9699 // CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9700 // CHECK19-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
9701 // CHECK19-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9702 // CHECK19:       .omp.final.then:
9703 // CHECK19-NEXT:    store i64 400, i64* [[IT]], align 8
9704 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9705 // CHECK19:       .omp.final.done:
9706 // CHECK19-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9707 // CHECK19-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
9708 // CHECK19-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
9709 // CHECK19:       .omp.linear.pu:
9710 // CHECK19-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN2]], align 4
9711 // CHECK19-NEXT:    store i32 [[TMP22]], i32* [[LIN_ADDR]], align 4
9712 // CHECK19-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A3]], align 4
9713 // CHECK19-NEXT:    store i32 [[TMP23]], i32* [[A_ADDR]], align 4
9714 // CHECK19-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
9715 // CHECK19:       .omp.linear.pu.done:
9716 // CHECK19-NEXT:    ret void
9717 //
9718 //
9719 // CHECK19-LABEL: define {{[^@]+}}@_Z7get_valv
9720 // CHECK19-SAME: () #[[ATTR3:[0-9]+]] {
9721 // CHECK19-NEXT:  entry:
9722 // CHECK19-NEXT:    ret i64 0
9723 //
9724 //
9725 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
9726 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
9727 // CHECK19-NEXT:  entry:
9728 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9729 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9730 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9731 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
9732 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9733 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9734 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9735 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
9736 // CHECK19-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
9737 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
9738 // CHECK19-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
9739 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
9740 // CHECK19-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
9741 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
9742 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
9743 // CHECK19-NEXT:    ret void
9744 //
9745 //
9746 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
9747 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR1]] {
9748 // CHECK19-NEXT:  entry:
9749 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9750 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9751 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9752 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
9753 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9754 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i16, align 2
9755 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9756 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9757 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9758 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9759 // CHECK19-NEXT:    [[IT:%.*]] = alloca i16, align 2
9760 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9761 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9762 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9763 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
9764 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
9765 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9766 // CHECK19-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
9767 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9768 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9769 // CHECK19-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9770 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9771 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9772 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9773 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
9774 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9775 // CHECK19:       cond.true:
9776 // CHECK19-NEXT:    br label [[COND_END:%.*]]
9777 // CHECK19:       cond.false:
9778 // CHECK19-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9779 // CHECK19-NEXT:    br label [[COND_END]]
9780 // CHECK19:       cond.end:
9781 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9782 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9783 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9784 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9785 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9786 // CHECK19:       omp.inner.for.cond:
9787 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9788 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
9789 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9790 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9791 // CHECK19:       omp.inner.for.body:
9792 // CHECK19-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9793 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
9794 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
9795 // CHECK19-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
9796 // CHECK19-NEXT:    store i16 [[CONV2]], i16* [[IT]], align 2, !llvm.access.group !21
9797 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !21
9798 // CHECK19-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
9799 // CHECK19-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !21
9800 // CHECK19-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !21
9801 // CHECK19-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
9802 // CHECK19-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
9803 // CHECK19-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
9804 // CHECK19-NEXT:    store i16 [[CONV6]], i16* [[CONV]], align 2, !llvm.access.group !21
9805 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9806 // CHECK19:       omp.body.continue:
9807 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9808 // CHECK19:       omp.inner.for.inc:
9809 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9810 // CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
9811 // CHECK19-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9812 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
9813 // CHECK19:       omp.inner.for.end:
9814 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9815 // CHECK19:       omp.loop.exit:
9816 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9817 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9818 // CHECK19-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9819 // CHECK19-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9820 // CHECK19:       .omp.final.then:
9821 // CHECK19-NEXT:    store i16 22, i16* [[IT]], align 2
9822 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9823 // CHECK19:       .omp.final.done:
9824 // CHECK19-NEXT:    ret void
9825 //
9826 //
9827 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
9828 // CHECK19-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
9829 // CHECK19-NEXT:  entry:
9830 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9831 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9832 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9833 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9834 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9835 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9836 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9837 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9838 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9839 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9840 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
9841 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
9842 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9843 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9844 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9845 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9846 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9847 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9848 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9849 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9850 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9851 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9852 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9853 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9854 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9855 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9856 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9857 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9858 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9859 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9860 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
9861 // CHECK19-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
9862 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
9863 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9864 // CHECK19-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9865 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9866 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
9867 // CHECK19-NEXT:    ret void
9868 //
9869 //
9870 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
9871 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
9872 // CHECK19-NEXT:  entry:
9873 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9874 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9875 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
9876 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
9877 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
9878 // CHECK19-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
9879 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
9880 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
9881 // CHECK19-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
9882 // CHECK19-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
9883 // CHECK19-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
9884 // CHECK19-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9885 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9886 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i8, align 1
9887 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9888 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9889 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9890 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9891 // CHECK19-NEXT:    [[IT:%.*]] = alloca i8, align 1
9892 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9893 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9894 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
9895 // CHECK19-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
9896 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
9897 // CHECK19-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
9898 // CHECK19-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
9899 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
9900 // CHECK19-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
9901 // CHECK19-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
9902 // CHECK19-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
9903 // CHECK19-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9904 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
9905 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
9906 // CHECK19-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
9907 // CHECK19-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
9908 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
9909 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
9910 // CHECK19-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
9911 // CHECK19-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
9912 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9913 // CHECK19-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
9914 // CHECK19-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9915 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9916 // CHECK19-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9917 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9918 // CHECK19-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
9919 // CHECK19-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
9920 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
9921 // CHECK19:       omp.dispatch.cond:
9922 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9923 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
9924 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9925 // CHECK19:       cond.true:
9926 // CHECK19-NEXT:    br label [[COND_END:%.*]]
9927 // CHECK19:       cond.false:
9928 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9929 // CHECK19-NEXT:    br label [[COND_END]]
9930 // CHECK19:       cond.end:
9931 // CHECK19-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
9932 // CHECK19-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9933 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9934 // CHECK19-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
9935 // CHECK19-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9936 // CHECK19-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9937 // CHECK19-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
9938 // CHECK19-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9939 // CHECK19:       omp.dispatch.body:
9940 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9941 // CHECK19:       omp.inner.for.cond:
9942 // CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9943 // CHECK19-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
9944 // CHECK19-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
9945 // CHECK19-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9946 // CHECK19:       omp.inner.for.body:
9947 // CHECK19-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9948 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
9949 // CHECK19-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
9950 // CHECK19-NEXT:    [[CONV:%.*]] = trunc i32 [[SUB]] to i8
9951 // CHECK19-NEXT:    store i8 [[CONV]], i8* [[IT]], align 1, !llvm.access.group !24
9952 // CHECK19-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !24
9953 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
9954 // CHECK19-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4, !llvm.access.group !24
9955 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
9956 // CHECK19-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !24
9957 // CHECK19-NEXT:    [[CONV7:%.*]] = fpext float [[TMP20]] to double
9958 // CHECK19-NEXT:    [[ADD8:%.*]] = fadd double [[CONV7]], 1.000000e+00
9959 // CHECK19-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
9960 // CHECK19-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !24
9961 // CHECK19-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
9962 // CHECK19-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !24
9963 // CHECK19-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
9964 // CHECK19-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
9965 // CHECK19-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
9966 // CHECK19-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !24
9967 // CHECK19-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
9968 // CHECK19-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
9969 // CHECK19-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !24
9970 // CHECK19-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
9971 // CHECK19-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !24
9972 // CHECK19-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
9973 // CHECK19-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
9974 // CHECK19-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
9975 // CHECK19-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !24
9976 // CHECK19-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
9977 // CHECK19-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !24
9978 // CHECK19-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
9979 // CHECK19-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !24
9980 // CHECK19-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
9981 // CHECK19-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !24
9982 // CHECK19-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
9983 // CHECK19-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !24
9984 // CHECK19-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
9985 // CHECK19-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
9986 // CHECK19-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
9987 // CHECK19-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !24
9988 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9989 // CHECK19:       omp.body.continue:
9990 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9991 // CHECK19:       omp.inner.for.inc:
9992 // CHECK19-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9993 // CHECK19-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
9994 // CHECK19-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9995 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
9996 // CHECK19:       omp.inner.for.end:
9997 // CHECK19-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
9998 // CHECK19:       omp.dispatch.inc:
9999 // CHECK19-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10000 // CHECK19-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10001 // CHECK19-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
10002 // CHECK19-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
10003 // CHECK19-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10004 // CHECK19-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10005 // CHECK19-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
10006 // CHECK19-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
10007 // CHECK19-NEXT:    br label [[OMP_DISPATCH_COND]]
10008 // CHECK19:       omp.dispatch.end:
10009 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
10010 // CHECK19-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10011 // CHECK19-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
10012 // CHECK19-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10013 // CHECK19:       .omp.final.then:
10014 // CHECK19-NEXT:    store i8 96, i8* [[IT]], align 1
10015 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10016 // CHECK19:       .omp.final.done:
10017 // CHECK19-NEXT:    ret void
10018 //
10019 //
10020 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
10021 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10022 // CHECK19-NEXT:  entry:
10023 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10024 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10025 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
10026 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10027 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10028 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10029 // CHECK19-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
10030 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10031 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10032 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
10033 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10034 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10035 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
10036 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10037 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
10038 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
10039 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
10040 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
10041 // CHECK19-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10042 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
10043 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10044 // CHECK19-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
10045 // CHECK19-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
10046 // CHECK19-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
10047 // CHECK19-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
10048 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
10049 // CHECK19-NEXT:    ret void
10050 //
10051 //
10052 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..4
10053 // 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:%.*]]) #[[ATTR1]] {
10054 // CHECK19-NEXT:  entry:
10055 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10056 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10057 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10058 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10059 // CHECK19-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
10060 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10061 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10062 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10063 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10064 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10065 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10066 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10067 // CHECK19-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
10068 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10069 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10070 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
10071 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10072 // CHECK19-NEXT:    ret void
10073 //
10074 //
10075 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l216
10076 // CHECK19-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]]) #[[ATTR0]] {
10077 // CHECK19-NEXT:  entry:
10078 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10079 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
10080 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
10081 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
10082 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
10083 // CHECK19-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
10084 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10085 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
10086 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
10087 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
10088 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
10089 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10090 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
10091 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
10092 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
10093 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[B_ADDR]], align 4
10094 // CHECK19-NEXT:    store i32 [[TMP4]], i32* [[B_CASTED]], align 4
10095 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_CASTED]], align 4
10096 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP0]], i32 [[TMP5]], i32 [[TMP1]], i32 [[TMP2]], i16* [[TMP3]])
10097 // CHECK19-NEXT:    ret void
10098 //
10099 //
10100 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..5
10101 // 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:%.*]]) #[[ATTR1]] {
10102 // CHECK19-NEXT:  entry:
10103 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10104 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10105 // CHECK19-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
10106 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
10107 // CHECK19-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
10108 // CHECK19-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
10109 // CHECK19-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
10110 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
10111 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i64, align 4
10112 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
10113 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
10114 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
10115 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10116 // CHECK19-NEXT:    [[IT:%.*]] = alloca i64, align 8
10117 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10118 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10119 // CHECK19-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
10120 // CHECK19-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
10121 // CHECK19-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
10122 // CHECK19-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
10123 // CHECK19-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
10124 // CHECK19-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
10125 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
10126 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
10127 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
10128 // CHECK19-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
10129 // CHECK19-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
10130 // CHECK19-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
10131 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10132 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10133 // CHECK19-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
10134 // CHECK19-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
10135 // CHECK19-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10136 // CHECK19-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP6]], 3
10137 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10138 // CHECK19:       cond.true:
10139 // CHECK19-NEXT:    br label [[COND_END:%.*]]
10140 // CHECK19:       cond.false:
10141 // CHECK19-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10142 // CHECK19-NEXT:    br label [[COND_END]]
10143 // CHECK19:       cond.end:
10144 // CHECK19-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
10145 // CHECK19-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
10146 // CHECK19-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
10147 // CHECK19-NEXT:    store i64 [[TMP8]], i64* [[DOTOMP_IV]], align 8
10148 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10149 // CHECK19:       omp.inner.for.cond:
10150 // CHECK19-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
10151 // CHECK19-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !27
10152 // CHECK19-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP9]], [[TMP10]]
10153 // CHECK19-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10154 // CHECK19:       omp.inner.for.body:
10155 // CHECK19-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
10156 // CHECK19-NEXT:    [[MUL:%.*]] = mul i64 [[TMP11]], 400
10157 // CHECK19-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
10158 // CHECK19-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !27
10159 // CHECK19-NEXT:    [[TMP12:%.*]] = load i32, i32* [[B_ADDR]], align 4, !llvm.access.group !27
10160 // CHECK19-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP12]] to double
10161 // CHECK19-NEXT:    [[ADD:%.*]] = fadd double [[CONV]], 1.500000e+00
10162 // CHECK19-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
10163 // CHECK19-NEXT:    store double [[ADD]], double* [[A]], align 4, !llvm.access.group !27
10164 // CHECK19-NEXT:    [[A4:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
10165 // CHECK19-NEXT:    [[TMP13:%.*]] = load double, double* [[A4]], align 4, !llvm.access.group !27
10166 // CHECK19-NEXT:    [[INC:%.*]] = fadd double [[TMP13]], 1.000000e+00
10167 // CHECK19-NEXT:    store double [[INC]], double* [[A4]], align 4, !llvm.access.group !27
10168 // CHECK19-NEXT:    [[CONV5:%.*]] = fptosi double [[INC]] to i16
10169 // CHECK19-NEXT:    [[TMP14:%.*]] = mul nsw i32 1, [[TMP2]]
10170 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP14]]
10171 // CHECK19-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
10172 // CHECK19-NEXT:    store i16 [[CONV5]], i16* [[ARRAYIDX6]], align 2, !llvm.access.group !27
10173 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10174 // CHECK19:       omp.body.continue:
10175 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10176 // CHECK19:       omp.inner.for.inc:
10177 // CHECK19-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
10178 // CHECK19-NEXT:    [[ADD7:%.*]] = add i64 [[TMP15]], 1
10179 // CHECK19-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
10180 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
10181 // CHECK19:       omp.inner.for.end:
10182 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10183 // CHECK19:       omp.loop.exit:
10184 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
10185 // CHECK19-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10186 // CHECK19-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
10187 // CHECK19-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10188 // CHECK19:       .omp.final.then:
10189 // CHECK19-NEXT:    store i64 400, i64* [[IT]], align 8
10190 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10191 // CHECK19:       .omp.final.done:
10192 // CHECK19-NEXT:    ret void
10193 //
10194 //
10195 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
10196 // CHECK19-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10197 // CHECK19-NEXT:  entry:
10198 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10199 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10200 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10201 // CHECK19-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
10202 // CHECK19-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
10203 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10204 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10205 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10206 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10207 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10208 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
10209 // CHECK19-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
10210 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
10211 // CHECK19-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
10212 // CHECK19-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
10213 // CHECK19-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
10214 // CHECK19-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
10215 // CHECK19-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
10216 // CHECK19-NEXT:    ret void
10217 //
10218 //
10219 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
10220 // 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:%.*]]) #[[ATTR1]] {
10221 // CHECK19-NEXT:  entry:
10222 // CHECK19-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
10223 // CHECK19-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
10224 // CHECK19-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
10225 // CHECK19-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
10226 // CHECK19-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
10227 // CHECK19-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
10228 // CHECK19-NEXT:    [[TMP:%.*]] = alloca i64, align 4
10229 // CHECK19-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
10230 // CHECK19-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
10231 // CHECK19-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
10232 // CHECK19-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10233 // CHECK19-NEXT:    [[I:%.*]] = alloca i64, align 8
10234 // CHECK19-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
10235 // CHECK19-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
10236 // CHECK19-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
10237 // CHECK19-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
10238 // CHECK19-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
10239 // CHECK19-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
10240 // CHECK19-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
10241 // CHECK19-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
10242 // CHECK19-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
10243 // CHECK19-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
10244 // CHECK19-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10245 // CHECK19-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
10246 // CHECK19-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10247 // CHECK19-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
10248 // CHECK19-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10249 // CHECK19-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
10250 // CHECK19-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10251 // CHECK19:       cond.true:
10252 // CHECK19-NEXT:    br label [[COND_END:%.*]]
10253 // CHECK19:       cond.false:
10254 // CHECK19-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10255 // CHECK19-NEXT:    br label [[COND_END]]
10256 // CHECK19:       cond.end:
10257 // CHECK19-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
10258 // CHECK19-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
10259 // CHECK19-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
10260 // CHECK19-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
10261 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10262 // CHECK19:       omp.inner.for.cond:
10263 // CHECK19-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
10264 // CHECK19-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !30
10265 // CHECK19-NEXT:    [[CMP1:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
10266 // CHECK19-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10267 // CHECK19:       omp.inner.for.body:
10268 // CHECK19-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
10269 // CHECK19-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
10270 // CHECK19-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
10271 // CHECK19-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !30
10272 // CHECK19-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !30
10273 // CHECK19-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
10274 // CHECK19-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4, !llvm.access.group !30
10275 // CHECK19-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !30
10276 // CHECK19-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
10277 // CHECK19-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
10278 // CHECK19-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
10279 // CHECK19-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2, !llvm.access.group !30
10280 // CHECK19-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
10281 // CHECK19-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !30
10282 // CHECK19-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
10283 // CHECK19-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !30
10284 // CHECK19-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10285 // CHECK19:       omp.body.continue:
10286 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10287 // CHECK19:       omp.inner.for.inc:
10288 // CHECK19-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
10289 // CHECK19-NEXT:    [[ADD7:%.*]] = add nsw i64 [[TMP12]], 1
10290 // CHECK19-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !30
10291 // CHECK19-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
10292 // CHECK19:       omp.inner.for.end:
10293 // CHECK19-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10294 // CHECK19:       omp.loop.exit:
10295 // CHECK19-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
10296 // CHECK19-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10297 // CHECK19-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
10298 // CHECK19-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10299 // CHECK19:       .omp.final.then:
10300 // CHECK19-NEXT:    store i64 11, i64* [[I]], align 8
10301 // CHECK19-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10302 // CHECK19:       .omp.final.done:
10303 // CHECK19-NEXT:    ret void
10304 //
10305 //
10306 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
10307 // CHECK21-SAME: () #[[ATTR0:[0-9]+]] {
10308 // CHECK21-NEXT:  entry:
10309 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
10310 // CHECK21-NEXT:    ret void
10311 //
10312 //
10313 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined.
10314 // CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
10315 // CHECK21-NEXT:  entry:
10316 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10317 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10318 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10319 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10320 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10321 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10322 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10323 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10324 // CHECK21-NEXT:    [[I:%.*]] = alloca i32, align 4
10325 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10326 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10327 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10328 // CHECK21-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
10329 // CHECK21-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10330 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10331 // CHECK21-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10332 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10333 // CHECK21-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10334 // CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10335 // CHECK21-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
10336 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10337 // CHECK21:       cond.true:
10338 // CHECK21-NEXT:    br label [[COND_END:%.*]]
10339 // CHECK21:       cond.false:
10340 // CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10341 // CHECK21-NEXT:    br label [[COND_END]]
10342 // CHECK21:       cond.end:
10343 // CHECK21-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10344 // CHECK21-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10345 // CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10346 // CHECK21-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10347 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10348 // CHECK21:       omp.inner.for.cond:
10349 // CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
10350 // CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
10351 // CHECK21-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10352 // CHECK21-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10353 // CHECK21:       omp.inner.for.body:
10354 // CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
10355 // CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
10356 // CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
10357 // CHECK21-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
10358 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10359 // CHECK21:       omp.body.continue:
10360 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10361 // CHECK21:       omp.inner.for.inc:
10362 // CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
10363 // CHECK21-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
10364 // CHECK21-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
10365 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
10366 // CHECK21:       omp.inner.for.end:
10367 // CHECK21-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10368 // CHECK21:       omp.loop.exit:
10369 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10370 // CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10371 // CHECK21-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
10372 // CHECK21-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10373 // CHECK21:       .omp.final.then:
10374 // CHECK21-NEXT:    store i32 33, i32* [[I]], align 4
10375 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10376 // CHECK21:       .omp.final.done:
10377 // CHECK21-NEXT:    ret void
10378 //
10379 //
10380 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
10381 // CHECK21-SAME: (i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR0]] {
10382 // CHECK21-NEXT:  entry:
10383 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10384 // CHECK21-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
10385 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10386 // CHECK21-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10387 // CHECK21-NEXT:    [[LIN_CASTED:%.*]] = alloca i64, align 8
10388 // CHECK21-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10389 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10390 // CHECK21-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
10391 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10392 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10393 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
10394 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10395 // CHECK21-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
10396 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10397 // CHECK21-NEXT:    store i16 [[TMP0]], i16* [[CONV3]], align 2
10398 // CHECK21-NEXT:    [[TMP1:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10399 // CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[CONV1]], align 4
10400 // CHECK21-NEXT:    [[CONV4:%.*]] = bitcast i64* [[LIN_CASTED]] to i32*
10401 // CHECK21-NEXT:    store i32 [[TMP2]], i32* [[CONV4]], align 4
10402 // CHECK21-NEXT:    [[TMP3:%.*]] = load i64, i64* [[LIN_CASTED]], align 8
10403 // CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[CONV2]], align 4
10404 // CHECK21-NEXT:    [[CONV5:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10405 // CHECK21-NEXT:    store i32 [[TMP4]], i32* [[CONV5]], align 4
10406 // CHECK21-NEXT:    [[TMP5:%.*]] = load i64, i64* [[A_CASTED]], align 8
10407 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]], i64 [[TMP5]])
10408 // CHECK21-NEXT:    ret void
10409 //
10410 //
10411 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..1
10412 // CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[LIN:%.*]], i64 noundef [[A:%.*]]) #[[ATTR1]] {
10413 // CHECK21-NEXT:  entry:
10414 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10415 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10416 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10417 // CHECK21-NEXT:    [[LIN_ADDR:%.*]] = alloca i64, align 8
10418 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10419 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
10420 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i64, align 8
10421 // CHECK21-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
10422 // CHECK21-NEXT:    [[DOTLINEAR_START3:%.*]] = alloca i32, align 4
10423 // CHECK21-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
10424 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
10425 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
10426 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
10427 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10428 // CHECK21-NEXT:    [[IT:%.*]] = alloca i64, align 8
10429 // CHECK21-NEXT:    [[LIN4:%.*]] = alloca i32, align 4
10430 // CHECK21-NEXT:    [[A5:%.*]] = alloca i32, align 4
10431 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10432 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10433 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10434 // CHECK21-NEXT:    store i64 [[LIN]], i64* [[LIN_ADDR]], align 8
10435 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10436 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10437 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[LIN_ADDR]] to i32*
10438 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10439 // CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV1]], align 4
10440 // CHECK21-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
10441 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV2]], align 4
10442 // CHECK21-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START3]], align 4
10443 // CHECK21-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv() #[[ATTR5:[0-9]+]]
10444 // CHECK21-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
10445 // CHECK21-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
10446 // CHECK21-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
10447 // CHECK21-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
10448 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10449 // CHECK21-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10450 // CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10451 // CHECK21-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
10452 // CHECK21-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
10453 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10454 // CHECK21-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
10455 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10456 // CHECK21:       cond.true:
10457 // CHECK21-NEXT:    br label [[COND_END:%.*]]
10458 // CHECK21:       cond.false:
10459 // CHECK21-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10460 // CHECK21-NEXT:    br label [[COND_END]]
10461 // CHECK21:       cond.end:
10462 // CHECK21-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10463 // CHECK21-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
10464 // CHECK21-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
10465 // CHECK21-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
10466 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10467 // CHECK21:       omp.inner.for.cond:
10468 // CHECK21-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10469 // CHECK21-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !17
10470 // CHECK21-NEXT:    [[CMP6:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
10471 // CHECK21-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10472 // CHECK21:       omp.inner.for.body:
10473 // CHECK21-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10474 // CHECK21-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
10475 // CHECK21-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
10476 // CHECK21-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !17
10477 // CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !17
10478 // CHECK21-NEXT:    [[CONV7:%.*]] = sext i32 [[TMP10]] to i64
10479 // CHECK21-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10480 // CHECK21-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !17
10481 // CHECK21-NEXT:    [[MUL8:%.*]] = mul i64 [[TMP11]], [[TMP12]]
10482 // CHECK21-NEXT:    [[ADD:%.*]] = add i64 [[CONV7]], [[MUL8]]
10483 // CHECK21-NEXT:    [[CONV9:%.*]] = trunc i64 [[ADD]] to i32
10484 // CHECK21-NEXT:    store i32 [[CONV9]], i32* [[LIN4]], align 4, !llvm.access.group !17
10485 // CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START3]], align 4, !llvm.access.group !17
10486 // CHECK21-NEXT:    [[CONV10:%.*]] = sext i32 [[TMP13]] to i64
10487 // CHECK21-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10488 // CHECK21-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !17
10489 // CHECK21-NEXT:    [[MUL11:%.*]] = mul i64 [[TMP14]], [[TMP15]]
10490 // CHECK21-NEXT:    [[ADD12:%.*]] = add i64 [[CONV10]], [[MUL11]]
10491 // CHECK21-NEXT:    [[CONV13:%.*]] = trunc i64 [[ADD12]] to i32
10492 // CHECK21-NEXT:    store i32 [[CONV13]], i32* [[A5]], align 4, !llvm.access.group !17
10493 // CHECK21-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !17
10494 // CHECK21-NEXT:    [[CONV14:%.*]] = sext i16 [[TMP16]] to i32
10495 // CHECK21-NEXT:    [[ADD15:%.*]] = add nsw i32 [[CONV14]], 1
10496 // CHECK21-NEXT:    [[CONV16:%.*]] = trunc i32 [[ADD15]] to i16
10497 // CHECK21-NEXT:    store i16 [[CONV16]], i16* [[CONV]], align 2, !llvm.access.group !17
10498 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10499 // CHECK21:       omp.body.continue:
10500 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10501 // CHECK21:       omp.inner.for.inc:
10502 // CHECK21-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10503 // CHECK21-NEXT:    [[ADD17:%.*]] = add i64 [[TMP17]], 1
10504 // CHECK21-NEXT:    store i64 [[ADD17]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !17
10505 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
10506 // CHECK21:       omp.inner.for.end:
10507 // CHECK21-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10508 // CHECK21:       omp.loop.exit:
10509 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10510 // CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10511 // CHECK21-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
10512 // CHECK21-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10513 // CHECK21:       .omp.final.then:
10514 // CHECK21-NEXT:    store i64 400, i64* [[IT]], align 8
10515 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10516 // CHECK21:       .omp.final.done:
10517 // CHECK21-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10518 // CHECK21-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
10519 // CHECK21-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
10520 // CHECK21:       .omp.linear.pu:
10521 // CHECK21-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN4]], align 4
10522 // CHECK21-NEXT:    store i32 [[TMP22]], i32* [[CONV1]], align 4
10523 // CHECK21-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A5]], align 4
10524 // CHECK21-NEXT:    store i32 [[TMP23]], i32* [[CONV2]], align 4
10525 // CHECK21-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
10526 // CHECK21:       .omp.linear.pu.done:
10527 // CHECK21-NEXT:    ret void
10528 //
10529 //
10530 // CHECK21-LABEL: define {{[^@]+}}@_Z7get_valv
10531 // CHECK21-SAME: () #[[ATTR3:[0-9]+]] {
10532 // CHECK21-NEXT:  entry:
10533 // CHECK21-NEXT:    ret i64 0
10534 //
10535 //
10536 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
10537 // CHECK21-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR0]] {
10538 // CHECK21-NEXT:  entry:
10539 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10540 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10541 // CHECK21-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10542 // CHECK21-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10543 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10544 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10545 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10546 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10547 // CHECK21-NEXT:    [[TMP0:%.*]] = load i32, i32* [[CONV]], align 4
10548 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10549 // CHECK21-NEXT:    store i32 [[TMP0]], i32* [[CONV2]], align 4
10550 // CHECK21-NEXT:    [[TMP1:%.*]] = load i64, i64* [[A_CASTED]], align 8
10551 // CHECK21-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV1]], align 2
10552 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10553 // CHECK21-NEXT:    store i16 [[TMP2]], i16* [[CONV3]], align 2
10554 // CHECK21-NEXT:    [[TMP3:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10555 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i64 [[TMP1]], i64 [[TMP3]])
10556 // CHECK21-NEXT:    ret void
10557 //
10558 //
10559 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..2
10560 // CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]]) #[[ATTR1]] {
10561 // CHECK21-NEXT:  entry:
10562 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10563 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10564 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10565 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10566 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10567 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i16, align 2
10568 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10569 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10570 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10571 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10572 // CHECK21-NEXT:    [[IT:%.*]] = alloca i16, align 2
10573 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10574 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10575 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10576 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10577 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10578 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10579 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10580 // CHECK21-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
10581 // CHECK21-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10582 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10583 // CHECK21-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10584 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10585 // CHECK21-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10586 // CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10587 // CHECK21-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
10588 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10589 // CHECK21:       cond.true:
10590 // CHECK21-NEXT:    br label [[COND_END:%.*]]
10591 // CHECK21:       cond.false:
10592 // CHECK21-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10593 // CHECK21-NEXT:    br label [[COND_END]]
10594 // CHECK21:       cond.end:
10595 // CHECK21-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10596 // CHECK21-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10597 // CHECK21-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10598 // CHECK21-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10599 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10600 // CHECK21:       omp.inner.for.cond:
10601 // CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
10602 // CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !20
10603 // CHECK21-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10604 // CHECK21-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10605 // CHECK21:       omp.inner.for.body:
10606 // CHECK21-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
10607 // CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
10608 // CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
10609 // CHECK21-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i16
10610 // CHECK21-NEXT:    store i16 [[CONV3]], i16* [[IT]], align 2, !llvm.access.group !20
10611 // CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !20
10612 // CHECK21-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
10613 // CHECK21-NEXT:    store i32 [[ADD4]], i32* [[CONV]], align 4, !llvm.access.group !20
10614 // CHECK21-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !20
10615 // CHECK21-NEXT:    [[CONV5:%.*]] = sext i16 [[TMP9]] to i32
10616 // CHECK21-NEXT:    [[ADD6:%.*]] = add nsw i32 [[CONV5]], 1
10617 // CHECK21-NEXT:    [[CONV7:%.*]] = trunc i32 [[ADD6]] to i16
10618 // CHECK21-NEXT:    store i16 [[CONV7]], i16* [[CONV1]], align 2, !llvm.access.group !20
10619 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10620 // CHECK21:       omp.body.continue:
10621 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10622 // CHECK21:       omp.inner.for.inc:
10623 // CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
10624 // CHECK21-NEXT:    [[ADD8:%.*]] = add nsw i32 [[TMP10]], 1
10625 // CHECK21-NEXT:    store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
10626 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
10627 // CHECK21:       omp.inner.for.end:
10628 // CHECK21-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10629 // CHECK21:       omp.loop.exit:
10630 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10631 // CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10632 // CHECK21-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10633 // CHECK21-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10634 // CHECK21:       .omp.final.then:
10635 // CHECK21-NEXT:    store i16 22, i16* [[IT]], align 2
10636 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10637 // CHECK21:       .omp.final.done:
10638 // CHECK21-NEXT:    ret void
10639 //
10640 //
10641 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
10642 // CHECK21-SAME: (i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
10643 // CHECK21-NEXT:  entry:
10644 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10645 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
10646 // CHECK21-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10647 // CHECK21-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
10648 // CHECK21-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
10649 // CHECK21-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10650 // CHECK21-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
10651 // CHECK21-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
10652 // CHECK21-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
10653 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10654 // CHECK21-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10655 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10656 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10657 // CHECK21-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
10658 // CHECK21-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10659 // CHECK21-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
10660 // CHECK21-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
10661 // CHECK21-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10662 // CHECK21-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
10663 // CHECK21-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
10664 // CHECK21-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
10665 // CHECK21-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10666 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10667 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
10668 // CHECK21-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10669 // CHECK21-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
10670 // CHECK21-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
10671 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10672 // CHECK21-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
10673 // CHECK21-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
10674 // CHECK21-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
10675 // CHECK21-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
10676 // CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV]], align 4
10677 // CHECK21-NEXT:    [[CONV6:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10678 // CHECK21-NEXT:    store i32 [[TMP8]], i32* [[CONV6]], align 4
10679 // CHECK21-NEXT:    [[TMP9:%.*]] = load i64, i64* [[A_CASTED]], align 8
10680 // CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[CONV5]], align 4
10681 // CHECK21-NEXT:    [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
10682 // CHECK21-NEXT:    store i32 [[TMP10]], i32* [[CONV7]], align 4
10683 // CHECK21-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
10684 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, [10 x float]*, i64, float*, [5 x [10 x double]]*, i64, i64, double*, %struct.TT*, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], [10 x float]* [[TMP0]], i64 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i64 [[TMP4]], i64 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i64 [[TMP11]])
10685 // CHECK21-NEXT:    ret void
10686 //
10687 //
10688 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..3
10689 // CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i64 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 8 dereferenceable(400) [[C:%.*]], i64 noundef [[VLA1:%.*]], i64 noundef [[VLA3:%.*]], double* noundef nonnull align 8 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 8 dereferenceable(16) [[D:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
10690 // CHECK21-NEXT:  entry:
10691 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10692 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10693 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10694 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 8
10695 // CHECK21-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10696 // CHECK21-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 8
10697 // CHECK21-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 8
10698 // CHECK21-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10699 // CHECK21-NEXT:    [[VLA_ADDR4:%.*]] = alloca i64, align 8
10700 // CHECK21-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 8
10701 // CHECK21-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 8
10702 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10703 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10704 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i8, align 1
10705 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10706 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10707 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10708 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10709 // CHECK21-NEXT:    [[IT:%.*]] = alloca i8, align 1
10710 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10711 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10712 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10713 // CHECK21-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 8
10714 // CHECK21-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10715 // CHECK21-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 8
10716 // CHECK21-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 8
10717 // CHECK21-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10718 // CHECK21-NEXT:    store i64 [[VLA3]], i64* [[VLA_ADDR4]], align 8
10719 // CHECK21-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 8
10720 // CHECK21-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 8
10721 // CHECK21-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10722 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10723 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 8
10724 // CHECK21-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10725 // CHECK21-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 8
10726 // CHECK21-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 8
10727 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10728 // CHECK21-NEXT:    [[TMP5:%.*]] = load i64, i64* [[VLA_ADDR4]], align 8
10729 // CHECK21-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 8
10730 // CHECK21-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 8
10731 // CHECK21-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
10732 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10733 // CHECK21-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
10734 // CHECK21-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10735 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10736 // CHECK21-NEXT:    [[TMP8:%.*]] = load i32, i32* [[CONV5]], align 4
10737 // CHECK21-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10738 // CHECK21-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
10739 // CHECK21-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
10740 // CHECK21-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
10741 // CHECK21:       omp.dispatch.cond:
10742 // CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10743 // CHECK21-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
10744 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10745 // CHECK21:       cond.true:
10746 // CHECK21-NEXT:    br label [[COND_END:%.*]]
10747 // CHECK21:       cond.false:
10748 // CHECK21-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10749 // CHECK21-NEXT:    br label [[COND_END]]
10750 // CHECK21:       cond.end:
10751 // CHECK21-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
10752 // CHECK21-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10753 // CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10754 // CHECK21-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
10755 // CHECK21-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10756 // CHECK21-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10757 // CHECK21-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10758 // CHECK21-NEXT:    br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
10759 // CHECK21:       omp.dispatch.body:
10760 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10761 // CHECK21:       omp.inner.for.cond:
10762 // CHECK21-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
10763 // CHECK21-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
10764 // CHECK21-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
10765 // CHECK21-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10766 // CHECK21:       omp.inner.for.body:
10767 // CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
10768 // CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
10769 // CHECK21-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
10770 // CHECK21-NEXT:    [[CONV8:%.*]] = trunc i32 [[SUB]] to i8
10771 // CHECK21-NEXT:    store i8 [[CONV8]], i8* [[IT]], align 1, !llvm.access.group !23
10772 // CHECK21-NEXT:    [[TMP19:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !23
10773 // CHECK21-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
10774 // CHECK21-NEXT:    store i32 [[ADD]], i32* [[CONV]], align 4, !llvm.access.group !23
10775 // CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i64 0, i64 2
10776 // CHECK21-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !23
10777 // CHECK21-NEXT:    [[CONV9:%.*]] = fpext float [[TMP20]] to double
10778 // CHECK21-NEXT:    [[ADD10:%.*]] = fadd double [[CONV9]], 1.000000e+00
10779 // CHECK21-NEXT:    [[CONV11:%.*]] = fptrunc double [[ADD10]] to float
10780 // CHECK21-NEXT:    store float [[CONV11]], float* [[ARRAYIDX]], align 4, !llvm.access.group !23
10781 // CHECK21-NEXT:    [[ARRAYIDX12:%.*]] = getelementptr inbounds float, float* [[TMP2]], i64 3
10782 // CHECK21-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX12]], align 4, !llvm.access.group !23
10783 // CHECK21-NEXT:    [[CONV13:%.*]] = fpext float [[TMP21]] to double
10784 // CHECK21-NEXT:    [[ADD14:%.*]] = fadd double [[CONV13]], 1.000000e+00
10785 // CHECK21-NEXT:    [[CONV15:%.*]] = fptrunc double [[ADD14]] to float
10786 // CHECK21-NEXT:    store float [[CONV15]], float* [[ARRAYIDX12]], align 4, !llvm.access.group !23
10787 // CHECK21-NEXT:    [[ARRAYIDX16:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i64 0, i64 1
10788 // CHECK21-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX16]], i64 0, i64 2
10789 // CHECK21-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX17]], align 8, !llvm.access.group !23
10790 // CHECK21-NEXT:    [[ADD18:%.*]] = fadd double [[TMP22]], 1.000000e+00
10791 // CHECK21-NEXT:    store double [[ADD18]], double* [[ARRAYIDX17]], align 8, !llvm.access.group !23
10792 // CHECK21-NEXT:    [[TMP23:%.*]] = mul nsw i64 1, [[TMP5]]
10793 // CHECK21-NEXT:    [[ARRAYIDX19:%.*]] = getelementptr inbounds double, double* [[TMP6]], i64 [[TMP23]]
10794 // CHECK21-NEXT:    [[ARRAYIDX20:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX19]], i64 3
10795 // CHECK21-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX20]], align 8, !llvm.access.group !23
10796 // CHECK21-NEXT:    [[ADD21:%.*]] = fadd double [[TMP24]], 1.000000e+00
10797 // CHECK21-NEXT:    store double [[ADD21]], double* [[ARRAYIDX20]], align 8, !llvm.access.group !23
10798 // CHECK21-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
10799 // CHECK21-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 8, !llvm.access.group !23
10800 // CHECK21-NEXT:    [[ADD22:%.*]] = add nsw i64 [[TMP25]], 1
10801 // CHECK21-NEXT:    store i64 [[ADD22]], i64* [[X]], align 8, !llvm.access.group !23
10802 // CHECK21-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
10803 // CHECK21-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 8, !llvm.access.group !23
10804 // CHECK21-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP26]] to i32
10805 // CHECK21-NEXT:    [[ADD24:%.*]] = add nsw i32 [[CONV23]], 1
10806 // CHECK21-NEXT:    [[CONV25:%.*]] = trunc i32 [[ADD24]] to i8
10807 // CHECK21-NEXT:    store i8 [[CONV25]], i8* [[Y]], align 8, !llvm.access.group !23
10808 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10809 // CHECK21:       omp.body.continue:
10810 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10811 // CHECK21:       omp.inner.for.inc:
10812 // CHECK21-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
10813 // CHECK21-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP27]], 1
10814 // CHECK21-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
10815 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
10816 // CHECK21:       omp.inner.for.end:
10817 // CHECK21-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
10818 // CHECK21:       omp.dispatch.inc:
10819 // CHECK21-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10820 // CHECK21-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10821 // CHECK21-NEXT:    [[ADD27:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
10822 // CHECK21-NEXT:    store i32 [[ADD27]], i32* [[DOTOMP_LB]], align 4
10823 // CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10824 // CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10825 // CHECK21-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
10826 // CHECK21-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_UB]], align 4
10827 // CHECK21-NEXT:    br label [[OMP_DISPATCH_COND]]
10828 // CHECK21:       omp.dispatch.end:
10829 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
10830 // CHECK21-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10831 // CHECK21-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
10832 // CHECK21-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10833 // CHECK21:       .omp.final.then:
10834 // CHECK21-NEXT:    store i8 96, i8* [[IT]], align 1
10835 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10836 // CHECK21:       .omp.final.done:
10837 // CHECK21-NEXT:    ret void
10838 //
10839 //
10840 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
10841 // CHECK21-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], i64 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
10842 // CHECK21-NEXT:  entry:
10843 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10844 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10845 // CHECK21-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
10846 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10847 // CHECK21-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
10848 // CHECK21-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
10849 // CHECK21-NEXT:    [[AAA_CASTED:%.*]] = alloca i64, align 8
10850 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10851 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10852 // CHECK21-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
10853 // CHECK21-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10854 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10855 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10856 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
10857 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10858 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
10859 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[A_CASTED]] to i32*
10860 // CHECK21-NEXT:    store i32 [[TMP1]], i32* [[CONV3]], align 4
10861 // CHECK21-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
10862 // CHECK21-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
10863 // CHECK21-NEXT:    [[CONV4:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
10864 // CHECK21-NEXT:    store i16 [[TMP3]], i16* [[CONV4]], align 2
10865 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
10866 // CHECK21-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV2]], align 1
10867 // CHECK21-NEXT:    [[CONV5:%.*]] = bitcast i64* [[AAA_CASTED]] to i8*
10868 // CHECK21-NEXT:    store i8 [[TMP5]], i8* [[CONV5]], align 1
10869 // CHECK21-NEXT:    [[TMP6:%.*]] = load i64, i64* [[AAA_CASTED]], align 8
10870 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [10 x i32]* [[TMP0]])
10871 // CHECK21-NEXT:    ret void
10872 //
10873 //
10874 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..4
10875 // CHECK21-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:%.*]]) #[[ATTR1]] {
10876 // CHECK21-NEXT:  entry:
10877 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10878 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10879 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
10880 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
10881 // CHECK21-NEXT:    [[AAA_ADDR:%.*]] = alloca i64, align 8
10882 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
10883 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10884 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10885 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10886 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10887 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
10888 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
10889 // CHECK21-NEXT:    store i64 [[AAA]], i64* [[AAA_ADDR]], align 8
10890 // CHECK21-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
10891 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
10892 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
10893 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[AAA_ADDR]] to i8*
10894 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
10895 // CHECK21-NEXT:    ret void
10896 //
10897 //
10898 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214
10899 // CHECK21-SAME: (%struct.S1* noundef [[THIS:%.*]], i64 noundef [[B:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
10900 // CHECK21-NEXT:  entry:
10901 // CHECK21-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10902 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
10903 // CHECK21-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10904 // CHECK21-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10905 // CHECK21-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
10906 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10907 // CHECK21-NEXT:    [[B_CASTED:%.*]] = alloca i64, align 8
10908 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10909 // CHECK21-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
10910 // CHECK21-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10911 // CHECK21-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
10912 // CHECK21-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10913 // CHECK21-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
10914 // CHECK21-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10915 // CHECK21-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10916 // CHECK21-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
10917 // CHECK21-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10918 // CHECK21-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10919 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
10920 // CHECK21-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10921 // CHECK21-NEXT:    [[TMP3:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10922 // CHECK21-NEXT:    [[TMP4:%.*]] = load i16*, i16** [[C_ADDR]], align 8
10923 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
10924 // CHECK21-NEXT:    [[TMP5:%.*]] = load i32, i32* [[CONV]], align 4
10925 // CHECK21-NEXT:    [[CONV4:%.*]] = bitcast i64* [[B_CASTED]] to i32*
10926 // CHECK21-NEXT:    store i32 [[TMP5]], i32* [[CONV4]], align 4
10927 // CHECK21-NEXT:    [[TMP6:%.*]] = load i64, i64* [[B_CASTED]], align 8
10928 // CHECK21-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV3]], align 1
10929 // CHECK21-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
10930 // CHECK21-NEXT:    [[CONV5:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
10931 // CHECK21-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
10932 // CHECK21-NEXT:    store i8 [[FROMBOOL]], i8* [[CONV5]], align 1
10933 // CHECK21-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
10934 // CHECK21-NEXT:    [[TMP9:%.*]] = load i8, i8* [[CONV3]], align 1
10935 // CHECK21-NEXT:    [[TOBOOL6:%.*]] = trunc i8 [[TMP9]] to i1
10936 // CHECK21-NEXT:    br i1 [[TOBOOL6]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10937 // CHECK21:       omp_if.then:
10938 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i64, i64, i64, i16*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i64 [[TMP6]], i64 [[TMP2]], i64 [[TMP3]], i16* [[TMP4]], i64 [[TMP8]])
10939 // CHECK21-NEXT:    br label [[OMP_IF_END:%.*]]
10940 // CHECK21:       omp_if.else:
10941 // CHECK21-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
10942 // CHECK21-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
10943 // CHECK21-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
10944 // CHECK21-NEXT:    call void @.omp_outlined..5(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i64 [[TMP6]], i64 [[TMP2]], i64 [[TMP3]], i16* [[TMP4]], i64 [[TMP8]]) #[[ATTR2:[0-9]+]]
10945 // CHECK21-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
10946 // CHECK21-NEXT:    br label [[OMP_IF_END]]
10947 // CHECK21:       omp_if.end:
10948 // CHECK21-NEXT:    ret void
10949 //
10950 //
10951 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..5
10952 // CHECK21-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:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
10953 // CHECK21-NEXT:  entry:
10954 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10955 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10956 // CHECK21-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 8
10957 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca i64, align 8
10958 // CHECK21-NEXT:    [[VLA_ADDR:%.*]] = alloca i64, align 8
10959 // CHECK21-NEXT:    [[VLA_ADDR2:%.*]] = alloca i64, align 8
10960 // CHECK21-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 8
10961 // CHECK21-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10962 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
10963 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i64, align 8
10964 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
10965 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
10966 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
10967 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10968 // CHECK21-NEXT:    [[IT:%.*]] = alloca i64, align 8
10969 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10970 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10971 // CHECK21-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 8
10972 // CHECK21-NEXT:    store i64 [[B]], i64* [[B_ADDR]], align 8
10973 // CHECK21-NEXT:    store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10974 // CHECK21-NEXT:    store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
10975 // CHECK21-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 8
10976 // CHECK21-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10977 // CHECK21-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 8
10978 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[B_ADDR]] to i32*
10979 // CHECK21-NEXT:    [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10980 // CHECK21-NEXT:    [[TMP2:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
10981 // CHECK21-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 8
10982 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
10983 // CHECK21-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
10984 // CHECK21-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
10985 // CHECK21-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
10986 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10987 // CHECK21-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV3]], align 1
10988 // CHECK21-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
10989 // CHECK21-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10990 // CHECK21:       omp_if.then:
10991 // CHECK21-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10992 // CHECK21-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
10993 // CHECK21-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
10994 // CHECK21-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
10995 // CHECK21-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP7]], 3
10996 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10997 // CHECK21:       cond.true:
10998 // CHECK21-NEXT:    br label [[COND_END:%.*]]
10999 // CHECK21:       cond.false:
11000 // CHECK21-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11001 // CHECK21-NEXT:    br label [[COND_END]]
11002 // CHECK21:       cond.end:
11003 // CHECK21-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
11004 // CHECK21-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
11005 // CHECK21-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11006 // CHECK21-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_IV]], align 8
11007 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11008 // CHECK21:       omp.inner.for.cond:
11009 // CHECK21-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
11010 // CHECK21-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !26
11011 // CHECK21-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP10]], [[TMP11]]
11012 // CHECK21-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11013 // CHECK21:       omp.inner.for.body:
11014 // CHECK21-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
11015 // CHECK21-NEXT:    [[MUL:%.*]] = mul i64 [[TMP12]], 400
11016 // CHECK21-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
11017 // CHECK21-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !26
11018 // CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !26
11019 // CHECK21-NEXT:    [[CONV5:%.*]] = sitofp i32 [[TMP13]] to double
11020 // CHECK21-NEXT:    [[ADD:%.*]] = fadd double [[CONV5]], 1.500000e+00
11021 // CHECK21-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
11022 // CHECK21-NEXT:    store double [[ADD]], double* [[A]], align 8, !nontemporal !27, !llvm.access.group !26
11023 // CHECK21-NEXT:    [[A6:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11024 // CHECK21-NEXT:    [[TMP14:%.*]] = load double, double* [[A6]], align 8, !nontemporal !27, !llvm.access.group !26
11025 // CHECK21-NEXT:    [[INC:%.*]] = fadd double [[TMP14]], 1.000000e+00
11026 // CHECK21-NEXT:    store double [[INC]], double* [[A6]], align 8, !nontemporal !27, !llvm.access.group !26
11027 // CHECK21-NEXT:    [[CONV7:%.*]] = fptosi double [[INC]] to i16
11028 // CHECK21-NEXT:    [[TMP15:%.*]] = mul nsw i64 1, [[TMP2]]
11029 // CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP15]]
11030 // CHECK21-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i64 1
11031 // CHECK21-NEXT:    store i16 [[CONV7]], i16* [[ARRAYIDX8]], align 2, !llvm.access.group !26
11032 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11033 // CHECK21:       omp.body.continue:
11034 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11035 // CHECK21:       omp.inner.for.inc:
11036 // CHECK21-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
11037 // CHECK21-NEXT:    [[ADD9:%.*]] = add i64 [[TMP16]], 1
11038 // CHECK21-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !26
11039 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
11040 // CHECK21:       omp.inner.for.end:
11041 // CHECK21-NEXT:    br label [[OMP_IF_END:%.*]]
11042 // CHECK21:       omp_if.else:
11043 // CHECK21-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11044 // CHECK21-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
11045 // CHECK21-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
11046 // CHECK21-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11047 // CHECK21-NEXT:    [[CMP10:%.*]] = icmp ugt i64 [[TMP19]], 3
11048 // CHECK21-NEXT:    br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
11049 // CHECK21:       cond.true11:
11050 // CHECK21-NEXT:    br label [[COND_END13:%.*]]
11051 // CHECK21:       cond.false12:
11052 // CHECK21-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11053 // CHECK21-NEXT:    br label [[COND_END13]]
11054 // CHECK21:       cond.end13:
11055 // CHECK21-NEXT:    [[COND14:%.*]] = phi i64 [ 3, [[COND_TRUE11]] ], [ [[TMP20]], [[COND_FALSE12]] ]
11056 // CHECK21-NEXT:    store i64 [[COND14]], i64* [[DOTOMP_UB]], align 8
11057 // CHECK21-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11058 // CHECK21-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV]], align 8
11059 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND15:%.*]]
11060 // CHECK21:       omp.inner.for.cond15:
11061 // CHECK21-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11062 // CHECK21-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11063 // CHECK21-NEXT:    [[CMP16:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
11064 // CHECK21-NEXT:    br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY17:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
11065 // CHECK21:       omp.inner.for.body17:
11066 // CHECK21-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11067 // CHECK21-NEXT:    [[MUL18:%.*]] = mul i64 [[TMP24]], 400
11068 // CHECK21-NEXT:    [[SUB19:%.*]] = sub i64 2000, [[MUL18]]
11069 // CHECK21-NEXT:    store i64 [[SUB19]], i64* [[IT]], align 8
11070 // CHECK21-NEXT:    [[TMP25:%.*]] = load i32, i32* [[CONV]], align 4
11071 // CHECK21-NEXT:    [[CONV20:%.*]] = sitofp i32 [[TMP25]] to double
11072 // CHECK21-NEXT:    [[ADD21:%.*]] = fadd double [[CONV20]], 1.500000e+00
11073 // CHECK21-NEXT:    [[A22:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11074 // CHECK21-NEXT:    store double [[ADD21]], double* [[A22]], align 8
11075 // CHECK21-NEXT:    [[A23:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11076 // CHECK21-NEXT:    [[TMP26:%.*]] = load double, double* [[A23]], align 8
11077 // CHECK21-NEXT:    [[INC24:%.*]] = fadd double [[TMP26]], 1.000000e+00
11078 // CHECK21-NEXT:    store double [[INC24]], double* [[A23]], align 8
11079 // CHECK21-NEXT:    [[CONV25:%.*]] = fptosi double [[INC24]] to i16
11080 // CHECK21-NEXT:    [[TMP27:%.*]] = mul nsw i64 1, [[TMP2]]
11081 // CHECK21-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i64 [[TMP27]]
11082 // CHECK21-NEXT:    [[ARRAYIDX27:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX26]], i64 1
11083 // CHECK21-NEXT:    store i16 [[CONV25]], i16* [[ARRAYIDX27]], align 2
11084 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE28:%.*]]
11085 // CHECK21:       omp.body.continue28:
11086 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC29:%.*]]
11087 // CHECK21:       omp.inner.for.inc29:
11088 // CHECK21-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11089 // CHECK21-NEXT:    [[ADD30:%.*]] = add i64 [[TMP28]], 1
11090 // CHECK21-NEXT:    store i64 [[ADD30]], i64* [[DOTOMP_IV]], align 8
11091 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND15]], !llvm.loop [[LOOP30:![0-9]+]]
11092 // CHECK21:       omp.inner.for.end31:
11093 // CHECK21-NEXT:    br label [[OMP_IF_END]]
11094 // CHECK21:       omp_if.end:
11095 // CHECK21-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11096 // CHECK21:       omp.loop.exit:
11097 // CHECK21-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11098 // CHECK21-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
11099 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
11100 // CHECK21-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11101 // CHECK21-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
11102 // CHECK21-NEXT:    br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11103 // CHECK21:       .omp.final.then:
11104 // CHECK21-NEXT:    store i64 400, i64* [[IT]], align 8
11105 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11106 // CHECK21:       .omp.final.done:
11107 // CHECK21-NEXT:    ret void
11108 //
11109 //
11110 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
11111 // CHECK21-SAME: (i64 noundef [[A:%.*]], i64 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11112 // CHECK21-NEXT:  entry:
11113 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11114 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11115 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11116 // CHECK21-NEXT:    [[A_CASTED:%.*]] = alloca i64, align 8
11117 // CHECK21-NEXT:    [[AA_CASTED:%.*]] = alloca i64, align 8
11118 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11119 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11120 // CHECK21-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11121 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11122 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11123 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11124 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
11125 // CHECK21-NEXT:    [[CONV2:%.*]] = bitcast i64* [[A_CASTED]] to i32*
11126 // CHECK21-NEXT:    store i32 [[TMP1]], i32* [[CONV2]], align 4
11127 // CHECK21-NEXT:    [[TMP2:%.*]] = load i64, i64* [[A_CASTED]], align 8
11128 // CHECK21-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV1]], align 2
11129 // CHECK21-NEXT:    [[CONV3:%.*]] = bitcast i64* [[AA_CASTED]] to i16*
11130 // CHECK21-NEXT:    store i16 [[TMP3]], i16* [[CONV3]], align 2
11131 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[AA_CASTED]], align 8
11132 // CHECK21-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP2]], i64 [[TMP4]], [10 x i32]* [[TMP0]])
11133 // CHECK21-NEXT:    ret void
11134 //
11135 //
11136 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined..6
11137 // CHECK21-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:%.*]]) #[[ATTR1]] {
11138 // CHECK21-NEXT:  entry:
11139 // CHECK21-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11140 // CHECK21-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11141 // CHECK21-NEXT:    [[A_ADDR:%.*]] = alloca i64, align 8
11142 // CHECK21-NEXT:    [[AA_ADDR:%.*]] = alloca i64, align 8
11143 // CHECK21-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 8
11144 // CHECK21-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
11145 // CHECK21-NEXT:    [[TMP:%.*]] = alloca i64, align 8
11146 // CHECK21-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
11147 // CHECK21-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
11148 // CHECK21-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
11149 // CHECK21-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11150 // CHECK21-NEXT:    [[I:%.*]] = alloca i64, align 8
11151 // CHECK21-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11152 // CHECK21-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11153 // CHECK21-NEXT:    store i64 [[A]], i64* [[A_ADDR]], align 8
11154 // CHECK21-NEXT:    store i64 [[AA]], i64* [[AA_ADDR]], align 8
11155 // CHECK21-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 8
11156 // CHECK21-NEXT:    [[CONV:%.*]] = bitcast i64* [[A_ADDR]] to i32*
11157 // CHECK21-NEXT:    [[CONV1:%.*]] = bitcast i64* [[AA_ADDR]] to i16*
11158 // CHECK21-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 8
11159 // CHECK21-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
11160 // CHECK21-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
11161 // CHECK21-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
11162 // CHECK21-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11163 // CHECK21-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11164 // CHECK21-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11165 // CHECK21-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
11166 // CHECK21-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11167 // CHECK21-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
11168 // CHECK21-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11169 // CHECK21:       cond.true:
11170 // CHECK21-NEXT:    br label [[COND_END:%.*]]
11171 // CHECK21:       cond.false:
11172 // CHECK21-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11173 // CHECK21-NEXT:    br label [[COND_END]]
11174 // CHECK21:       cond.end:
11175 // CHECK21-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
11176 // CHECK21-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
11177 // CHECK21-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11178 // CHECK21-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
11179 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11180 // CHECK21:       omp.inner.for.cond:
11181 // CHECK21-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !32
11182 // CHECK21-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !32
11183 // CHECK21-NEXT:    [[CMP2:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
11184 // CHECK21-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11185 // CHECK21:       omp.inner.for.body:
11186 // CHECK21-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !32
11187 // CHECK21-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
11188 // CHECK21-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
11189 // CHECK21-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !32
11190 // CHECK21-NEXT:    [[TMP9:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !32
11191 // CHECK21-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP9]], 1
11192 // CHECK21-NEXT:    store i32 [[ADD3]], i32* [[CONV]], align 4, !llvm.access.group !32
11193 // CHECK21-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV1]], align 2, !llvm.access.group !32
11194 // CHECK21-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP10]] to i32
11195 // CHECK21-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
11196 // CHECK21-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
11197 // CHECK21-NEXT:    store i16 [[CONV6]], i16* [[CONV1]], align 2, !llvm.access.group !32
11198 // CHECK21-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 2
11199 // CHECK21-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
11200 // CHECK21-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP11]], 1
11201 // CHECK21-NEXT:    store i32 [[ADD7]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !32
11202 // CHECK21-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11203 // CHECK21:       omp.body.continue:
11204 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11205 // CHECK21:       omp.inner.for.inc:
11206 // CHECK21-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !32
11207 // CHECK21-NEXT:    [[ADD8:%.*]] = add nsw i64 [[TMP12]], 1
11208 // CHECK21-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !32
11209 // CHECK21-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
11210 // CHECK21:       omp.inner.for.end:
11211 // CHECK21-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11212 // CHECK21:       omp.loop.exit:
11213 // CHECK21-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
11214 // CHECK21-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11215 // CHECK21-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
11216 // CHECK21-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11217 // CHECK21:       .omp.final.then:
11218 // CHECK21-NEXT:    store i64 11, i64* [[I]], align 8
11219 // CHECK21-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11220 // CHECK21:       .omp.final.done:
11221 // CHECK21-NEXT:    ret void
11222 //
11223 //
11224 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l96
11225 // CHECK23-SAME: () #[[ATTR0:[0-9]+]] {
11226 // CHECK23-NEXT:  entry:
11227 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
11228 // CHECK23-NEXT:    ret void
11229 //
11230 //
11231 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined.
11232 // CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1:[0-9]+]] {
11233 // CHECK23-NEXT:  entry:
11234 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11235 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11236 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11237 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11238 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11239 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11240 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11241 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11242 // CHECK23-NEXT:    [[I:%.*]] = alloca i32, align 4
11243 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11244 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11245 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11246 // CHECK23-NEXT:    store i32 5, i32* [[DOTOMP_UB]], align 4
11247 // CHECK23-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11248 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11249 // CHECK23-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11250 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11251 // CHECK23-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11252 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11253 // CHECK23-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 5
11254 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11255 // CHECK23:       cond.true:
11256 // CHECK23-NEXT:    br label [[COND_END:%.*]]
11257 // CHECK23:       cond.false:
11258 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11259 // CHECK23-NEXT:    br label [[COND_END]]
11260 // CHECK23:       cond.end:
11261 // CHECK23-NEXT:    [[COND:%.*]] = phi i32 [ 5, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11262 // CHECK23-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11263 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11264 // CHECK23-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11265 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11266 // CHECK23:       omp.inner.for.cond:
11267 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
11268 // CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
11269 // CHECK23-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11270 // CHECK23-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11271 // CHECK23:       omp.inner.for.body:
11272 // CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
11273 // CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 5
11274 // CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 3, [[MUL]]
11275 // CHECK23-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
11276 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11277 // CHECK23:       omp.body.continue:
11278 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11279 // CHECK23:       omp.inner.for.inc:
11280 // CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
11281 // CHECK23-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP8]], 1
11282 // CHECK23-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
11283 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
11284 // CHECK23:       omp.inner.for.end:
11285 // CHECK23-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11286 // CHECK23:       omp.loop.exit:
11287 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11288 // CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11289 // CHECK23-NEXT:    [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0
11290 // CHECK23-NEXT:    br i1 [[TMP10]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11291 // CHECK23:       .omp.final.then:
11292 // CHECK23-NEXT:    store i32 33, i32* [[I]], align 4
11293 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11294 // CHECK23:       .omp.final.done:
11295 // CHECK23-NEXT:    ret void
11296 //
11297 //
11298 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l108
11299 // CHECK23-SAME: (i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR0]] {
11300 // CHECK23-NEXT:  entry:
11301 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11302 // CHECK23-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
11303 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11304 // CHECK23-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11305 // CHECK23-NEXT:    [[LIN_CASTED:%.*]] = alloca i32, align 4
11306 // CHECK23-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11307 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11308 // CHECK23-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
11309 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11310 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11311 // CHECK23-NEXT:    [[TMP0:%.*]] = load i16, i16* [[CONV]], align 2
11312 // CHECK23-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11313 // CHECK23-NEXT:    store i16 [[TMP0]], i16* [[CONV1]], align 2
11314 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11315 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
11316 // CHECK23-NEXT:    store i32 [[TMP2]], i32* [[LIN_CASTED]], align 4
11317 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[LIN_CASTED]], align 4
11318 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[A_ADDR]], align 4
11319 // CHECK23-NEXT:    store i32 [[TMP4]], i32* [[A_CASTED]], align 4
11320 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[A_CASTED]], align 4
11321 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]], i32 [[TMP5]])
11322 // CHECK23-NEXT:    ret void
11323 //
11324 //
11325 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..1
11326 // CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[LIN:%.*]], i32 noundef [[A:%.*]]) #[[ATTR1]] {
11327 // CHECK23-NEXT:  entry:
11328 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11329 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11330 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11331 // CHECK23-NEXT:    [[LIN_ADDR:%.*]] = alloca i32, align 4
11332 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11333 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
11334 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i64, align 4
11335 // CHECK23-NEXT:    [[DOTLINEAR_START:%.*]] = alloca i32, align 4
11336 // CHECK23-NEXT:    [[DOTLINEAR_START1:%.*]] = alloca i32, align 4
11337 // CHECK23-NEXT:    [[DOTLINEAR_STEP:%.*]] = alloca i64, align 8
11338 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
11339 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
11340 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
11341 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11342 // CHECK23-NEXT:    [[IT:%.*]] = alloca i64, align 8
11343 // CHECK23-NEXT:    [[LIN2:%.*]] = alloca i32, align 4
11344 // CHECK23-NEXT:    [[A3:%.*]] = alloca i32, align 4
11345 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11346 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11347 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11348 // CHECK23-NEXT:    store i32 [[LIN]], i32* [[LIN_ADDR]], align 4
11349 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11350 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11351 // CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[LIN_ADDR]], align 4
11352 // CHECK23-NEXT:    store i32 [[TMP0]], i32* [[DOTLINEAR_START]], align 4
11353 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11354 // CHECK23-NEXT:    store i32 [[TMP1]], i32* [[DOTLINEAR_START1]], align 4
11355 // CHECK23-NEXT:    [[CALL:%.*]] = call noundef i64 @_Z7get_valv() #[[ATTR5:[0-9]+]]
11356 // CHECK23-NEXT:    store i64 [[CALL]], i64* [[DOTLINEAR_STEP]], align 8
11357 // CHECK23-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
11358 // CHECK23-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
11359 // CHECK23-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
11360 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11361 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11362 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
11363 // CHECK23-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP3]])
11364 // CHECK23-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
11365 // CHECK23-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11366 // CHECK23-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP4]], 3
11367 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11368 // CHECK23:       cond.true:
11369 // CHECK23-NEXT:    br label [[COND_END:%.*]]
11370 // CHECK23:       cond.false:
11371 // CHECK23-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11372 // CHECK23-NEXT:    br label [[COND_END]]
11373 // CHECK23:       cond.end:
11374 // CHECK23-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
11375 // CHECK23-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
11376 // CHECK23-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11377 // CHECK23-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_IV]], align 8
11378 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11379 // CHECK23:       omp.inner.for.cond:
11380 // CHECK23-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11381 // CHECK23-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !18
11382 // CHECK23-NEXT:    [[CMP4:%.*]] = icmp ule i64 [[TMP7]], [[TMP8]]
11383 // CHECK23-NEXT:    br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11384 // CHECK23:       omp.inner.for.body:
11385 // CHECK23-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11386 // CHECK23-NEXT:    [[MUL:%.*]] = mul i64 [[TMP9]], 400
11387 // CHECK23-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
11388 // CHECK23-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !18
11389 // CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTLINEAR_START]], align 4, !llvm.access.group !18
11390 // CHECK23-NEXT:    [[CONV5:%.*]] = sext i32 [[TMP10]] to i64
11391 // CHECK23-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11392 // CHECK23-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !18
11393 // CHECK23-NEXT:    [[MUL6:%.*]] = mul i64 [[TMP11]], [[TMP12]]
11394 // CHECK23-NEXT:    [[ADD:%.*]] = add i64 [[CONV5]], [[MUL6]]
11395 // CHECK23-NEXT:    [[CONV7:%.*]] = trunc i64 [[ADD]] to i32
11396 // CHECK23-NEXT:    store i32 [[CONV7]], i32* [[LIN2]], align 4, !llvm.access.group !18
11397 // CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTLINEAR_START1]], align 4, !llvm.access.group !18
11398 // CHECK23-NEXT:    [[CONV8:%.*]] = sext i32 [[TMP13]] to i64
11399 // CHECK23-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11400 // CHECK23-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTLINEAR_STEP]], align 8, !llvm.access.group !18
11401 // CHECK23-NEXT:    [[MUL9:%.*]] = mul i64 [[TMP14]], [[TMP15]]
11402 // CHECK23-NEXT:    [[ADD10:%.*]] = add i64 [[CONV8]], [[MUL9]]
11403 // CHECK23-NEXT:    [[CONV11:%.*]] = trunc i64 [[ADD10]] to i32
11404 // CHECK23-NEXT:    store i32 [[CONV11]], i32* [[A3]], align 4, !llvm.access.group !18
11405 // CHECK23-NEXT:    [[TMP16:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !18
11406 // CHECK23-NEXT:    [[CONV12:%.*]] = sext i16 [[TMP16]] to i32
11407 // CHECK23-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV12]], 1
11408 // CHECK23-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i16
11409 // CHECK23-NEXT:    store i16 [[CONV14]], i16* [[CONV]], align 2, !llvm.access.group !18
11410 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11411 // CHECK23:       omp.body.continue:
11412 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11413 // CHECK23:       omp.inner.for.inc:
11414 // CHECK23-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11415 // CHECK23-NEXT:    [[ADD15:%.*]] = add i64 [[TMP17]], 1
11416 // CHECK23-NEXT:    store i64 [[ADD15]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !18
11417 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
11418 // CHECK23:       omp.inner.for.end:
11419 // CHECK23-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11420 // CHECK23:       omp.loop.exit:
11421 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
11422 // CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11423 // CHECK23-NEXT:    [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
11424 // CHECK23-NEXT:    br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11425 // CHECK23:       .omp.final.then:
11426 // CHECK23-NEXT:    store i64 400, i64* [[IT]], align 8
11427 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11428 // CHECK23:       .omp.final.done:
11429 // CHECK23-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11430 // CHECK23-NEXT:    [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
11431 // CHECK23-NEXT:    br i1 [[TMP21]], label [[DOTOMP_LINEAR_PU:%.*]], label [[DOTOMP_LINEAR_PU_DONE:%.*]]
11432 // CHECK23:       .omp.linear.pu:
11433 // CHECK23-NEXT:    [[TMP22:%.*]] = load i32, i32* [[LIN2]], align 4
11434 // CHECK23-NEXT:    store i32 [[TMP22]], i32* [[LIN_ADDR]], align 4
11435 // CHECK23-NEXT:    [[TMP23:%.*]] = load i32, i32* [[A3]], align 4
11436 // CHECK23-NEXT:    store i32 [[TMP23]], i32* [[A_ADDR]], align 4
11437 // CHECK23-NEXT:    br label [[DOTOMP_LINEAR_PU_DONE]]
11438 // CHECK23:       .omp.linear.pu.done:
11439 // CHECK23-NEXT:    ret void
11440 //
11441 //
11442 // CHECK23-LABEL: define {{[^@]+}}@_Z7get_valv
11443 // CHECK23-SAME: () #[[ATTR3:[0-9]+]] {
11444 // CHECK23-NEXT:  entry:
11445 // CHECK23-NEXT:    ret i64 0
11446 //
11447 //
11448 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l116
11449 // CHECK23-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR0]] {
11450 // CHECK23-NEXT:  entry:
11451 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11452 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11453 // CHECK23-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11454 // CHECK23-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11455 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11456 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11457 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11458 // CHECK23-NEXT:    [[TMP0:%.*]] = load i32, i32* [[A_ADDR]], align 4
11459 // CHECK23-NEXT:    store i32 [[TMP0]], i32* [[A_CASTED]], align 4
11460 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_CASTED]], align 4
11461 // CHECK23-NEXT:    [[TMP2:%.*]] = load i16, i16* [[CONV]], align 2
11462 // CHECK23-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11463 // CHECK23-NEXT:    store i16 [[TMP2]], i16* [[CONV1]], align 2
11464 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11465 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32 [[TMP1]], i32 [[TMP3]])
11466 // CHECK23-NEXT:    ret void
11467 //
11468 //
11469 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..2
11470 // CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]]) #[[ATTR1]] {
11471 // CHECK23-NEXT:  entry:
11472 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11473 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11474 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11475 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11476 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11477 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i16, align 2
11478 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11479 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11480 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11481 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11482 // CHECK23-NEXT:    [[IT:%.*]] = alloca i16, align 2
11483 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11484 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11485 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11486 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11487 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11488 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11489 // CHECK23-NEXT:    store i32 3, i32* [[DOTOMP_UB]], align 4
11490 // CHECK23-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11491 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11492 // CHECK23-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11493 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11494 // CHECK23-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11495 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11496 // CHECK23-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 3
11497 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11498 // CHECK23:       cond.true:
11499 // CHECK23-NEXT:    br label [[COND_END:%.*]]
11500 // CHECK23:       cond.false:
11501 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11502 // CHECK23-NEXT:    br label [[COND_END]]
11503 // CHECK23:       cond.end:
11504 // CHECK23-NEXT:    [[COND:%.*]] = phi i32 [ 3, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11505 // CHECK23-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11506 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11507 // CHECK23-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11508 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11509 // CHECK23:       omp.inner.for.cond:
11510 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
11511 // CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
11512 // CHECK23-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11513 // CHECK23-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11514 // CHECK23:       omp.inner.for.body:
11515 // CHECK23-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
11516 // CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP7]], 4
11517 // CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 6, [[MUL]]
11518 // CHECK23-NEXT:    [[CONV2:%.*]] = trunc i32 [[ADD]] to i16
11519 // CHECK23-NEXT:    store i16 [[CONV2]], i16* [[IT]], align 2, !llvm.access.group !21
11520 // CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !21
11521 // CHECK23-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP8]], 1
11522 // CHECK23-NEXT:    store i32 [[ADD3]], i32* [[A_ADDR]], align 4, !llvm.access.group !21
11523 // CHECK23-NEXT:    [[TMP9:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !21
11524 // CHECK23-NEXT:    [[CONV4:%.*]] = sext i16 [[TMP9]] to i32
11525 // CHECK23-NEXT:    [[ADD5:%.*]] = add nsw i32 [[CONV4]], 1
11526 // CHECK23-NEXT:    [[CONV6:%.*]] = trunc i32 [[ADD5]] to i16
11527 // CHECK23-NEXT:    store i16 [[CONV6]], i16* [[CONV]], align 2, !llvm.access.group !21
11528 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11529 // CHECK23:       omp.body.continue:
11530 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11531 // CHECK23:       omp.inner.for.inc:
11532 // CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
11533 // CHECK23-NEXT:    [[ADD7:%.*]] = add nsw i32 [[TMP10]], 1
11534 // CHECK23-NEXT:    store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
11535 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
11536 // CHECK23:       omp.inner.for.end:
11537 // CHECK23-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11538 // CHECK23:       omp.loop.exit:
11539 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11540 // CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11541 // CHECK23-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
11542 // CHECK23-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11543 // CHECK23:       .omp.final.then:
11544 // CHECK23-NEXT:    store i16 22, i16* [[IT]], align 2
11545 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11546 // CHECK23:       .omp.final.done:
11547 // CHECK23-NEXT:    ret void
11548 //
11549 //
11550 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z3fooi_l140
11551 // CHECK23-SAME: (i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
11552 // CHECK23-NEXT:  entry:
11553 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11554 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11555 // CHECK23-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11556 // CHECK23-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11557 // CHECK23-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11558 // CHECK23-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11559 // CHECK23-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11560 // CHECK23-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11561 // CHECK23-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11562 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
11563 // CHECK23-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11564 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
11565 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11566 // CHECK23-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
11567 // CHECK23-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11568 // CHECK23-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
11569 // CHECK23-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
11570 // CHECK23-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11571 // CHECK23-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
11572 // CHECK23-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
11573 // CHECK23-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
11574 // CHECK23-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11575 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
11576 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11577 // CHECK23-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
11578 // CHECK23-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
11579 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11580 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
11581 // CHECK23-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
11582 // CHECK23-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
11583 // CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[A_ADDR]], align 4
11584 // CHECK23-NEXT:    store i32 [[TMP8]], i32* [[A_CASTED]], align 4
11585 // CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_CASTED]], align 4
11586 // CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11587 // CHECK23-NEXT:    store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
11588 // CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
11589 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 10, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, [10 x float]*, i32, float*, [5 x [10 x double]]*, i32, i32, double*, %struct.TT*, i32)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], [10 x float]* [[TMP0]], i32 [[TMP1]], float* [[TMP2]], [5 x [10 x double]]* [[TMP3]], i32 [[TMP4]], i32 [[TMP5]], double* [[TMP6]], %struct.TT* [[TMP7]], i32 [[TMP11]])
11590 // CHECK23-NEXT:    ret void
11591 //
11592 //
11593 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..3
11594 // CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[A:%.*]], [10 x float]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]], i32 noundef [[VLA:%.*]], float* noundef nonnull align 4 dereferenceable(4) [[BN:%.*]], [5 x [10 x double]]* noundef nonnull align 4 dereferenceable(400) [[C:%.*]], i32 noundef [[VLA1:%.*]], i32 noundef [[VLA3:%.*]], double* noundef nonnull align 4 dereferenceable(8) [[CN:%.*]], %struct.TT* noundef nonnull align 4 dereferenceable(12) [[D:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
11595 // CHECK23-NEXT:  entry:
11596 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11597 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11598 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11599 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x float]*, align 4
11600 // CHECK23-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11601 // CHECK23-NEXT:    [[BN_ADDR:%.*]] = alloca float*, align 4
11602 // CHECK23-NEXT:    [[C_ADDR:%.*]] = alloca [5 x [10 x double]]*, align 4
11603 // CHECK23-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11604 // CHECK23-NEXT:    [[VLA_ADDR4:%.*]] = alloca i32, align 4
11605 // CHECK23-NEXT:    [[CN_ADDR:%.*]] = alloca double*, align 4
11606 // CHECK23-NEXT:    [[D_ADDR:%.*]] = alloca %struct.TT*, align 4
11607 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
11608 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11609 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i8, align 1
11610 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11611 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11612 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11613 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11614 // CHECK23-NEXT:    [[IT:%.*]] = alloca i8, align 1
11615 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11616 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11617 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11618 // CHECK23-NEXT:    store [10 x float]* [[B]], [10 x float]** [[B_ADDR]], align 4
11619 // CHECK23-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11620 // CHECK23-NEXT:    store float* [[BN]], float** [[BN_ADDR]], align 4
11621 // CHECK23-NEXT:    store [5 x [10 x double]]* [[C]], [5 x [10 x double]]** [[C_ADDR]], align 4
11622 // CHECK23-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11623 // CHECK23-NEXT:    store i32 [[VLA3]], i32* [[VLA_ADDR4]], align 4
11624 // CHECK23-NEXT:    store double* [[CN]], double** [[CN_ADDR]], align 4
11625 // CHECK23-NEXT:    store %struct.TT* [[D]], %struct.TT** [[D_ADDR]], align 4
11626 // CHECK23-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11627 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x float]*, [10 x float]** [[B_ADDR]], align 4
11628 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11629 // CHECK23-NEXT:    [[TMP2:%.*]] = load float*, float** [[BN_ADDR]], align 4
11630 // CHECK23-NEXT:    [[TMP3:%.*]] = load [5 x [10 x double]]*, [5 x [10 x double]]** [[C_ADDR]], align 4
11631 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11632 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[VLA_ADDR4]], align 4
11633 // CHECK23-NEXT:    [[TMP6:%.*]] = load double*, double** [[CN_ADDR]], align 4
11634 // CHECK23-NEXT:    [[TMP7:%.*]] = load %struct.TT*, %struct.TT** [[D_ADDR]], align 4
11635 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11636 // CHECK23-NEXT:    store i32 25, i32* [[DOTOMP_UB]], align 4
11637 // CHECK23-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11638 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11639 // CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11640 // CHECK23-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11641 // CHECK23-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
11642 // CHECK23-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP8]])
11643 // CHECK23-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
11644 // CHECK23:       omp.dispatch.cond:
11645 // CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11646 // CHECK23-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 25
11647 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11648 // CHECK23:       cond.true:
11649 // CHECK23-NEXT:    br label [[COND_END:%.*]]
11650 // CHECK23:       cond.false:
11651 // CHECK23-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11652 // CHECK23-NEXT:    br label [[COND_END]]
11653 // CHECK23:       cond.end:
11654 // CHECK23-NEXT:    [[COND:%.*]] = phi i32 [ 25, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
11655 // CHECK23-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11656 // CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11657 // CHECK23-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
11658 // CHECK23-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11659 // CHECK23-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11660 // CHECK23-NEXT:    [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
11661 // CHECK23-NEXT:    br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
11662 // CHECK23:       omp.dispatch.body:
11663 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11664 // CHECK23:       omp.inner.for.cond:
11665 // CHECK23-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
11666 // CHECK23-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
11667 // CHECK23-NEXT:    [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
11668 // CHECK23-NEXT:    br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11669 // CHECK23:       omp.inner.for.body:
11670 // CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
11671 // CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
11672 // CHECK23-NEXT:    [[SUB:%.*]] = sub nsw i32 122, [[MUL]]
11673 // CHECK23-NEXT:    [[CONV:%.*]] = trunc i32 [[SUB]] to i8
11674 // CHECK23-NEXT:    store i8 [[CONV]], i8* [[IT]], align 1, !llvm.access.group !24
11675 // CHECK23-NEXT:    [[TMP19:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !24
11676 // CHECK23-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], 1
11677 // CHECK23-NEXT:    store i32 [[ADD]], i32* [[A_ADDR]], align 4, !llvm.access.group !24
11678 // CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x float], [10 x float]* [[TMP0]], i32 0, i32 2
11679 // CHECK23-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !24
11680 // CHECK23-NEXT:    [[CONV7:%.*]] = fpext float [[TMP20]] to double
11681 // CHECK23-NEXT:    [[ADD8:%.*]] = fadd double [[CONV7]], 1.000000e+00
11682 // CHECK23-NEXT:    [[CONV9:%.*]] = fptrunc double [[ADD8]] to float
11683 // CHECK23-NEXT:    store float [[CONV9]], float* [[ARRAYIDX]], align 4, !llvm.access.group !24
11684 // CHECK23-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 3
11685 // CHECK23-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !24
11686 // CHECK23-NEXT:    [[CONV11:%.*]] = fpext float [[TMP21]] to double
11687 // CHECK23-NEXT:    [[ADD12:%.*]] = fadd double [[CONV11]], 1.000000e+00
11688 // CHECK23-NEXT:    [[CONV13:%.*]] = fptrunc double [[ADD12]] to float
11689 // CHECK23-NEXT:    store float [[CONV13]], float* [[ARRAYIDX10]], align 4, !llvm.access.group !24
11690 // CHECK23-NEXT:    [[ARRAYIDX14:%.*]] = getelementptr inbounds [5 x [10 x double]], [5 x [10 x double]]* [[TMP3]], i32 0, i32 1
11691 // CHECK23-NEXT:    [[ARRAYIDX15:%.*]] = getelementptr inbounds [10 x double], [10 x double]* [[ARRAYIDX14]], i32 0, i32 2
11692 // CHECK23-NEXT:    [[TMP22:%.*]] = load double, double* [[ARRAYIDX15]], align 8, !llvm.access.group !24
11693 // CHECK23-NEXT:    [[ADD16:%.*]] = fadd double [[TMP22]], 1.000000e+00
11694 // CHECK23-NEXT:    store double [[ADD16]], double* [[ARRAYIDX15]], align 8, !llvm.access.group !24
11695 // CHECK23-NEXT:    [[TMP23:%.*]] = mul nsw i32 1, [[TMP5]]
11696 // CHECK23-NEXT:    [[ARRAYIDX17:%.*]] = getelementptr inbounds double, double* [[TMP6]], i32 [[TMP23]]
11697 // CHECK23-NEXT:    [[ARRAYIDX18:%.*]] = getelementptr inbounds double, double* [[ARRAYIDX17]], i32 3
11698 // CHECK23-NEXT:    [[TMP24:%.*]] = load double, double* [[ARRAYIDX18]], align 8, !llvm.access.group !24
11699 // CHECK23-NEXT:    [[ADD19:%.*]] = fadd double [[TMP24]], 1.000000e+00
11700 // CHECK23-NEXT:    store double [[ADD19]], double* [[ARRAYIDX18]], align 8, !llvm.access.group !24
11701 // CHECK23-NEXT:    [[X:%.*]] = getelementptr inbounds [[STRUCT_TT:%.*]], %struct.TT* [[TMP7]], i32 0, i32 0
11702 // CHECK23-NEXT:    [[TMP25:%.*]] = load i64, i64* [[X]], align 4, !llvm.access.group !24
11703 // CHECK23-NEXT:    [[ADD20:%.*]] = add nsw i64 [[TMP25]], 1
11704 // CHECK23-NEXT:    store i64 [[ADD20]], i64* [[X]], align 4, !llvm.access.group !24
11705 // CHECK23-NEXT:    [[Y:%.*]] = getelementptr inbounds [[STRUCT_TT]], %struct.TT* [[TMP7]], i32 0, i32 1
11706 // CHECK23-NEXT:    [[TMP26:%.*]] = load i8, i8* [[Y]], align 4, !llvm.access.group !24
11707 // CHECK23-NEXT:    [[CONV21:%.*]] = sext i8 [[TMP26]] to i32
11708 // CHECK23-NEXT:    [[ADD22:%.*]] = add nsw i32 [[CONV21]], 1
11709 // CHECK23-NEXT:    [[CONV23:%.*]] = trunc i32 [[ADD22]] to i8
11710 // CHECK23-NEXT:    store i8 [[CONV23]], i8* [[Y]], align 4, !llvm.access.group !24
11711 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11712 // CHECK23:       omp.body.continue:
11713 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11714 // CHECK23:       omp.inner.for.inc:
11715 // CHECK23-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
11716 // CHECK23-NEXT:    [[ADD24:%.*]] = add nsw i32 [[TMP27]], 1
11717 // CHECK23-NEXT:    store i32 [[ADD24]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
11718 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
11719 // CHECK23:       omp.inner.for.end:
11720 // CHECK23-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
11721 // CHECK23:       omp.dispatch.inc:
11722 // CHECK23-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11723 // CHECK23-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11724 // CHECK23-NEXT:    [[ADD25:%.*]] = add nsw i32 [[TMP28]], [[TMP29]]
11725 // CHECK23-NEXT:    store i32 [[ADD25]], i32* [[DOTOMP_LB]], align 4
11726 // CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11727 // CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11728 // CHECK23-NEXT:    [[ADD26:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
11729 // CHECK23-NEXT:    store i32 [[ADD26]], i32* [[DOTOMP_UB]], align 4
11730 // CHECK23-NEXT:    br label [[OMP_DISPATCH_COND]]
11731 // CHECK23:       omp.dispatch.end:
11732 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
11733 // CHECK23-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11734 // CHECK23-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
11735 // CHECK23-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11736 // CHECK23:       .omp.final.then:
11737 // CHECK23-NEXT:    store i8 96, i8* [[IT]], align 1
11738 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11739 // CHECK23:       .omp.final.done:
11740 // CHECK23-NEXT:    ret void
11741 //
11742 //
11743 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZL7fstatici_l195
11744 // CHECK23-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], i32 noundef [[AAA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
11745 // CHECK23-NEXT:  entry:
11746 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11747 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11748 // CHECK23-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
11749 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11750 // CHECK23-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
11751 // CHECK23-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
11752 // CHECK23-NEXT:    [[AAA_CASTED:%.*]] = alloca i32, align 4
11753 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11754 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11755 // CHECK23-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
11756 // CHECK23-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11757 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11758 // CHECK23-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
11759 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11760 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
11761 // CHECK23-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
11762 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
11763 // CHECK23-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
11764 // CHECK23-NEXT:    [[CONV2:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
11765 // CHECK23-NEXT:    store i16 [[TMP3]], i16* [[CONV2]], align 2
11766 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
11767 // CHECK23-NEXT:    [[TMP5:%.*]] = load i8, i8* [[CONV1]], align 1
11768 // CHECK23-NEXT:    [[CONV3:%.*]] = bitcast i32* [[AAA_CASTED]] to i8*
11769 // CHECK23-NEXT:    store i8 [[TMP5]], i8* [[CONV3]], align 1
11770 // CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[AAA_CASTED]], align 4
11771 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, [10 x i32]*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [10 x i32]* [[TMP0]])
11772 // CHECK23-NEXT:    ret void
11773 //
11774 //
11775 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..4
11776 // CHECK23-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:%.*]]) #[[ATTR1]] {
11777 // CHECK23-NEXT:  entry:
11778 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11779 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11780 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
11781 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
11782 // CHECK23-NEXT:    [[AAA_ADDR:%.*]] = alloca i32, align 4
11783 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
11784 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11785 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11786 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11787 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11788 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
11789 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
11790 // CHECK23-NEXT:    store i32 [[AAA]], i32* [[AAA_ADDR]], align 4
11791 // CHECK23-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
11792 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
11793 // CHECK23-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AAA_ADDR]] to i8*
11794 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
11795 // CHECK23-NEXT:    ret void
11796 //
11797 //
11798 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2S12r1Ei_l214
11799 // CHECK23-SAME: (%struct.S1* noundef [[THIS:%.*]], i32 noundef [[B:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i16* noundef nonnull align 2 dereferenceable(2) [[C:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR0]] {
11800 // CHECK23-NEXT:  entry:
11801 // CHECK23-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11802 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
11803 // CHECK23-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11804 // CHECK23-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11805 // CHECK23-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
11806 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
11807 // CHECK23-NEXT:    [[B_CASTED:%.*]] = alloca i32, align 4
11808 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
11809 // CHECK23-NEXT:    [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4
11810 // CHECK23-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
11811 // CHECK23-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
11812 // CHECK23-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11813 // CHECK23-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
11814 // CHECK23-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11815 // CHECK23-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11816 // CHECK23-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
11817 // CHECK23-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11818 // CHECK23-NEXT:    [[TMP1:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11819 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11820 // CHECK23-NEXT:    [[TMP3:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11821 // CHECK23-NEXT:    [[TMP4:%.*]] = load i16*, i16** [[C_ADDR]], align 4
11822 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
11823 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32, i32* [[B_ADDR]], align 4
11824 // CHECK23-NEXT:    store i32 [[TMP5]], i32* [[B_CASTED]], align 4
11825 // CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[B_CASTED]], align 4
11826 // CHECK23-NEXT:    [[TMP7:%.*]] = load i8, i8* [[CONV]], align 1
11827 // CHECK23-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP7]] to i1
11828 // CHECK23-NEXT:    [[CONV3:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
11829 // CHECK23-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
11830 // CHECK23-NEXT:    store i8 [[FROMBOOL]], i8* [[CONV3]], align 1
11831 // CHECK23-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
11832 // CHECK23-NEXT:    [[TMP9:%.*]] = load i8, i8* [[CONV]], align 1
11833 // CHECK23-NEXT:    [[TOBOOL4:%.*]] = trunc i8 [[TMP9]] to i1
11834 // CHECK23-NEXT:    br i1 [[TOBOOL4]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11835 // CHECK23:       omp_if.then:
11836 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.S1*, i32, i32, i32, i16*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.S1* [[TMP1]], i32 [[TMP6]], i32 [[TMP2]], i32 [[TMP3]], i16* [[TMP4]], i32 [[TMP8]])
11837 // CHECK23-NEXT:    br label [[OMP_IF_END:%.*]]
11838 // CHECK23:       omp_if.else:
11839 // CHECK23-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
11840 // CHECK23-NEXT:    store i32 [[TMP0]], i32* [[DOTTHREADID_TEMP_]], align 4
11841 // CHECK23-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
11842 // CHECK23-NEXT:    call void @.omp_outlined..5(i32* [[DOTTHREADID_TEMP_]], i32* [[DOTBOUND_ZERO_ADDR]], %struct.S1* [[TMP1]], i32 [[TMP6]], i32 [[TMP2]], i32 [[TMP3]], i16* [[TMP4]], i32 [[TMP8]]) #[[ATTR2:[0-9]+]]
11843 // CHECK23-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]])
11844 // CHECK23-NEXT:    br label [[OMP_IF_END]]
11845 // CHECK23:       omp_if.end:
11846 // CHECK23-NEXT:    ret void
11847 //
11848 //
11849 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..5
11850 // CHECK23-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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
11851 // CHECK23-NEXT:  entry:
11852 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
11853 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
11854 // CHECK23-NEXT:    [[THIS_ADDR:%.*]] = alloca %struct.S1*, align 4
11855 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4
11856 // CHECK23-NEXT:    [[VLA_ADDR:%.*]] = alloca i32, align 4
11857 // CHECK23-NEXT:    [[VLA_ADDR2:%.*]] = alloca i32, align 4
11858 // CHECK23-NEXT:    [[C_ADDR:%.*]] = alloca i16*, align 4
11859 // CHECK23-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
11860 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
11861 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i64, align 4
11862 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
11863 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
11864 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
11865 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11866 // CHECK23-NEXT:    [[IT:%.*]] = alloca i64, align 8
11867 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
11868 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
11869 // CHECK23-NEXT:    store %struct.S1* [[THIS]], %struct.S1** [[THIS_ADDR]], align 4
11870 // CHECK23-NEXT:    store i32 [[B]], i32* [[B_ADDR]], align 4
11871 // CHECK23-NEXT:    store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
11872 // CHECK23-NEXT:    store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
11873 // CHECK23-NEXT:    store i16* [[C]], i16** [[C_ADDR]], align 4
11874 // CHECK23-NEXT:    store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
11875 // CHECK23-NEXT:    [[TMP0:%.*]] = load %struct.S1*, %struct.S1** [[THIS_ADDR]], align 4
11876 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
11877 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
11878 // CHECK23-NEXT:    [[TMP3:%.*]] = load i16*, i16** [[C_ADDR]], align 4
11879 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
11880 // CHECK23-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
11881 // CHECK23-NEXT:    store i64 3, i64* [[DOTOMP_UB]], align 8
11882 // CHECK23-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
11883 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11884 // CHECK23-NEXT:    [[TMP4:%.*]] = load i8, i8* [[CONV]], align 1
11885 // CHECK23-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP4]] to i1
11886 // CHECK23-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11887 // CHECK23:       omp_if.then:
11888 // CHECK23-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11889 // CHECK23-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
11890 // CHECK23-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP6]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
11891 // CHECK23-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11892 // CHECK23-NEXT:    [[CMP:%.*]] = icmp ugt i64 [[TMP7]], 3
11893 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11894 // CHECK23:       cond.true:
11895 // CHECK23-NEXT:    br label [[COND_END:%.*]]
11896 // CHECK23:       cond.false:
11897 // CHECK23-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11898 // CHECK23-NEXT:    br label [[COND_END]]
11899 // CHECK23:       cond.end:
11900 // CHECK23-NEXT:    [[COND:%.*]] = phi i64 [ 3, [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
11901 // CHECK23-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
11902 // CHECK23-NEXT:    [[TMP9:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11903 // CHECK23-NEXT:    store i64 [[TMP9]], i64* [[DOTOMP_IV]], align 8
11904 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11905 // CHECK23:       omp.inner.for.cond:
11906 // CHECK23-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
11907 // CHECK23-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !27
11908 // CHECK23-NEXT:    [[CMP3:%.*]] = icmp ule i64 [[TMP10]], [[TMP11]]
11909 // CHECK23-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11910 // CHECK23:       omp.inner.for.body:
11911 // CHECK23-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
11912 // CHECK23-NEXT:    [[MUL:%.*]] = mul i64 [[TMP12]], 400
11913 // CHECK23-NEXT:    [[SUB:%.*]] = sub i64 2000, [[MUL]]
11914 // CHECK23-NEXT:    store i64 [[SUB]], i64* [[IT]], align 8, !llvm.access.group !27
11915 // CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[B_ADDR]], align 4, !llvm.access.group !27
11916 // CHECK23-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP13]] to double
11917 // CHECK23-NEXT:    [[ADD:%.*]] = fadd double [[CONV4]], 1.500000e+00
11918 // CHECK23-NEXT:    [[A:%.*]] = getelementptr inbounds [[STRUCT_S1:%.*]], %struct.S1* [[TMP0]], i32 0, i32 0
11919 // CHECK23-NEXT:    store double [[ADD]], double* [[A]], align 4, !nontemporal !28, !llvm.access.group !27
11920 // CHECK23-NEXT:    [[A5:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11921 // CHECK23-NEXT:    [[TMP14:%.*]] = load double, double* [[A5]], align 4, !nontemporal !28, !llvm.access.group !27
11922 // CHECK23-NEXT:    [[INC:%.*]] = fadd double [[TMP14]], 1.000000e+00
11923 // CHECK23-NEXT:    store double [[INC]], double* [[A5]], align 4, !nontemporal !28, !llvm.access.group !27
11924 // CHECK23-NEXT:    [[CONV6:%.*]] = fptosi double [[INC]] to i16
11925 // CHECK23-NEXT:    [[TMP15:%.*]] = mul nsw i32 1, [[TMP2]]
11926 // CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP15]]
11927 // CHECK23-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX]], i32 1
11928 // CHECK23-NEXT:    store i16 [[CONV6]], i16* [[ARRAYIDX7]], align 2, !llvm.access.group !27
11929 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11930 // CHECK23:       omp.body.continue:
11931 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11932 // CHECK23:       omp.inner.for.inc:
11933 // CHECK23-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
11934 // CHECK23-NEXT:    [[ADD8:%.*]] = add i64 [[TMP16]], 1
11935 // CHECK23-NEXT:    store i64 [[ADD8]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !27
11936 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
11937 // CHECK23:       omp.inner.for.end:
11938 // CHECK23-NEXT:    br label [[OMP_IF_END:%.*]]
11939 // CHECK23:       omp_if.else:
11940 // CHECK23-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11941 // CHECK23-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
11942 // CHECK23-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1]], i32 [[TMP18]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
11943 // CHECK23-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11944 // CHECK23-NEXT:    [[CMP9:%.*]] = icmp ugt i64 [[TMP19]], 3
11945 // CHECK23-NEXT:    br i1 [[CMP9]], label [[COND_TRUE10:%.*]], label [[COND_FALSE11:%.*]]
11946 // CHECK23:       cond.true10:
11947 // CHECK23-NEXT:    br label [[COND_END12:%.*]]
11948 // CHECK23:       cond.false11:
11949 // CHECK23-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11950 // CHECK23-NEXT:    br label [[COND_END12]]
11951 // CHECK23:       cond.end12:
11952 // CHECK23-NEXT:    [[COND13:%.*]] = phi i64 [ 3, [[COND_TRUE10]] ], [ [[TMP20]], [[COND_FALSE11]] ]
11953 // CHECK23-NEXT:    store i64 [[COND13]], i64* [[DOTOMP_UB]], align 8
11954 // CHECK23-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
11955 // CHECK23-NEXT:    store i64 [[TMP21]], i64* [[DOTOMP_IV]], align 8
11956 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND14:%.*]]
11957 // CHECK23:       omp.inner.for.cond14:
11958 // CHECK23-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11959 // CHECK23-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
11960 // CHECK23-NEXT:    [[CMP15:%.*]] = icmp ule i64 [[TMP22]], [[TMP23]]
11961 // CHECK23-NEXT:    br i1 [[CMP15]], label [[OMP_INNER_FOR_BODY16:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
11962 // CHECK23:       omp.inner.for.body16:
11963 // CHECK23-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11964 // CHECK23-NEXT:    [[MUL17:%.*]] = mul i64 [[TMP24]], 400
11965 // CHECK23-NEXT:    [[SUB18:%.*]] = sub i64 2000, [[MUL17]]
11966 // CHECK23-NEXT:    store i64 [[SUB18]], i64* [[IT]], align 8
11967 // CHECK23-NEXT:    [[TMP25:%.*]] = load i32, i32* [[B_ADDR]], align 4
11968 // CHECK23-NEXT:    [[CONV19:%.*]] = sitofp i32 [[TMP25]] to double
11969 // CHECK23-NEXT:    [[ADD20:%.*]] = fadd double [[CONV19]], 1.500000e+00
11970 // CHECK23-NEXT:    [[A21:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11971 // CHECK23-NEXT:    store double [[ADD20]], double* [[A21]], align 4
11972 // CHECK23-NEXT:    [[A22:%.*]] = getelementptr inbounds [[STRUCT_S1]], %struct.S1* [[TMP0]], i32 0, i32 0
11973 // CHECK23-NEXT:    [[TMP26:%.*]] = load double, double* [[A22]], align 4
11974 // CHECK23-NEXT:    [[INC23:%.*]] = fadd double [[TMP26]], 1.000000e+00
11975 // CHECK23-NEXT:    store double [[INC23]], double* [[A22]], align 4
11976 // CHECK23-NEXT:    [[CONV24:%.*]] = fptosi double [[INC23]] to i16
11977 // CHECK23-NEXT:    [[TMP27:%.*]] = mul nsw i32 1, [[TMP2]]
11978 // CHECK23-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds i16, i16* [[TMP3]], i32 [[TMP27]]
11979 // CHECK23-NEXT:    [[ARRAYIDX26:%.*]] = getelementptr inbounds i16, i16* [[ARRAYIDX25]], i32 1
11980 // CHECK23-NEXT:    store i16 [[CONV24]], i16* [[ARRAYIDX26]], align 2
11981 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
11982 // CHECK23:       omp.body.continue27:
11983 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
11984 // CHECK23:       omp.inner.for.inc28:
11985 // CHECK23-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
11986 // CHECK23-NEXT:    [[ADD29:%.*]] = add i64 [[TMP28]], 1
11987 // CHECK23-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8
11988 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND14]], !llvm.loop [[LOOP31:![0-9]+]]
11989 // CHECK23:       omp.inner.for.end30:
11990 // CHECK23-NEXT:    br label [[OMP_IF_END]]
11991 // CHECK23:       omp_if.end:
11992 // CHECK23-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11993 // CHECK23:       omp.loop.exit:
11994 // CHECK23-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
11995 // CHECK23-NEXT:    [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
11996 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
11997 // CHECK23-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11998 // CHECK23-NEXT:    [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
11999 // CHECK23-NEXT:    br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12000 // CHECK23:       .omp.final.then:
12001 // CHECK23-NEXT:    store i64 400, i64* [[IT]], align 8
12002 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12003 // CHECK23:       .omp.final.done:
12004 // CHECK23-NEXT:    ret void
12005 //
12006 //
12007 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9ftemplateIiET_i_l178
12008 // CHECK23-SAME: (i32 noundef [[A:%.*]], i32 noundef [[AA:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[B:%.*]]) #[[ATTR0]] {
12009 // CHECK23-NEXT:  entry:
12010 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12011 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12012 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
12013 // CHECK23-NEXT:    [[A_CASTED:%.*]] = alloca i32, align 4
12014 // CHECK23-NEXT:    [[AA_CASTED:%.*]] = alloca i32, align 4
12015 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12016 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12017 // CHECK23-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
12018 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12019 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
12020 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32, i32* [[A_ADDR]], align 4
12021 // CHECK23-NEXT:    store i32 [[TMP1]], i32* [[A_CASTED]], align 4
12022 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[A_CASTED]], align 4
12023 // CHECK23-NEXT:    [[TMP3:%.*]] = load i16, i16* [[CONV]], align 2
12024 // CHECK23-NEXT:    [[CONV1:%.*]] = bitcast i32* [[AA_CASTED]] to i16*
12025 // CHECK23-NEXT:    store i16 [[TMP3]], i16* [[CONV1]], align 2
12026 // CHECK23-NEXT:    [[TMP4:%.*]] = load i32, i32* [[AA_CASTED]], align 4
12027 // CHECK23-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP2]], i32 [[TMP4]], [10 x i32]* [[TMP0]])
12028 // CHECK23-NEXT:    ret void
12029 //
12030 //
12031 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined..6
12032 // CHECK23-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:%.*]]) #[[ATTR1]] {
12033 // CHECK23-NEXT:  entry:
12034 // CHECK23-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12035 // CHECK23-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12036 // CHECK23-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
12037 // CHECK23-NEXT:    [[AA_ADDR:%.*]] = alloca i32, align 4
12038 // CHECK23-NEXT:    [[B_ADDR:%.*]] = alloca [10 x i32]*, align 4
12039 // CHECK23-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
12040 // CHECK23-NEXT:    [[TMP:%.*]] = alloca i64, align 4
12041 // CHECK23-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
12042 // CHECK23-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
12043 // CHECK23-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
12044 // CHECK23-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12045 // CHECK23-NEXT:    [[I:%.*]] = alloca i64, align 8
12046 // CHECK23-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12047 // CHECK23-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12048 // CHECK23-NEXT:    store i32 [[A]], i32* [[A_ADDR]], align 4
12049 // CHECK23-NEXT:    store i32 [[AA]], i32* [[AA_ADDR]], align 4
12050 // CHECK23-NEXT:    store [10 x i32]* [[B]], [10 x i32]** [[B_ADDR]], align 4
12051 // CHECK23-NEXT:    [[CONV:%.*]] = bitcast i32* [[AA_ADDR]] to i16*
12052 // CHECK23-NEXT:    [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[B_ADDR]], align 4
12053 // CHECK23-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
12054 // CHECK23-NEXT:    store i64 6, i64* [[DOTOMP_UB]], align 8
12055 // CHECK23-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
12056 // CHECK23-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12057 // CHECK23-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12058 // CHECK23-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12059 // CHECK23-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
12060 // CHECK23-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
12061 // CHECK23-NEXT:    [[CMP:%.*]] = icmp sgt i64 [[TMP3]], 6
12062 // CHECK23-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12063 // CHECK23:       cond.true:
12064 // CHECK23-NEXT:    br label [[COND_END:%.*]]
12065 // CHECK23:       cond.false:
12066 // CHECK23-NEXT:    [[TMP4:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
12067 // CHECK23-NEXT:    br label [[COND_END]]
12068 // CHECK23:       cond.end:
12069 // CHECK23-NEXT:    [[COND:%.*]] = phi i64 [ 6, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
12070 // CHECK23-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
12071 // CHECK23-NEXT:    [[TMP5:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
12072 // CHECK23-NEXT:    store i64 [[TMP5]], i64* [[DOTOMP_IV]], align 8
12073 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12074 // CHECK23:       omp.inner.for.cond:
12075 // CHECK23-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !33
12076 // CHECK23-NEXT:    [[TMP7:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !33
12077 // CHECK23-NEXT:    [[CMP1:%.*]] = icmp sle i64 [[TMP6]], [[TMP7]]
12078 // CHECK23-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12079 // CHECK23:       omp.inner.for.body:
12080 // CHECK23-NEXT:    [[TMP8:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !33
12081 // CHECK23-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP8]], 3
12082 // CHECK23-NEXT:    [[ADD:%.*]] = add nsw i64 -10, [[MUL]]
12083 // CHECK23-NEXT:    store i64 [[ADD]], i64* [[I]], align 8, !llvm.access.group !33
12084 // CHECK23-NEXT:    [[TMP9:%.*]] = load i32, i32* [[A_ADDR]], align 4, !llvm.access.group !33
12085 // CHECK23-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP9]], 1
12086 // CHECK23-NEXT:    store i32 [[ADD2]], i32* [[A_ADDR]], align 4, !llvm.access.group !33
12087 // CHECK23-NEXT:    [[TMP10:%.*]] = load i16, i16* [[CONV]], align 2, !llvm.access.group !33
12088 // CHECK23-NEXT:    [[CONV3:%.*]] = sext i16 [[TMP10]] to i32
12089 // CHECK23-NEXT:    [[ADD4:%.*]] = add nsw i32 [[CONV3]], 1
12090 // CHECK23-NEXT:    [[CONV5:%.*]] = trunc i32 [[ADD4]] to i16
12091 // CHECK23-NEXT:    store i16 [[CONV5]], i16* [[CONV]], align 2, !llvm.access.group !33
12092 // CHECK23-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 2
12093 // CHECK23-NEXT:    [[TMP11:%.*]] = load i32, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
12094 // CHECK23-NEXT:    [[ADD6:%.*]] = add nsw i32 [[TMP11]], 1
12095 // CHECK23-NEXT:    store i32 [[ADD6]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
12096 // CHECK23-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12097 // CHECK23:       omp.body.continue:
12098 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12099 // CHECK23:       omp.inner.for.inc:
12100 // CHECK23-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !33
12101 // CHECK23-NEXT:    [[ADD7:%.*]] = add nsw i64 [[TMP12]], 1
12102 // CHECK23-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !33
12103 // CHECK23-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
12104 // CHECK23:       omp.inner.for.end:
12105 // CHECK23-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12106 // CHECK23:       omp.loop.exit:
12107 // CHECK23-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
12108 // CHECK23-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12109 // CHECK23-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
12110 // CHECK23-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12111 // CHECK23:       .omp.final.then:
12112 // CHECK23-NEXT:    store i64 11, i64* [[I]], align 8
12113 // CHECK23-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12114 // CHECK23:       .omp.final.done:
12115 // CHECK23-NEXT:    ret void
12116 //
12117