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 // expected-no-diagnostics
3 #ifndef HEADER
4 #define HEADER
5 // Test host codegen.
6 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
7 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
8 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
9 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
10 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
11 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
12
13 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
14 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
15 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
16 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK7
17 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
18 // RUN: %clang_cc1 -no-opaque-pointers -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7
19 #ifdef CK1
20
21 int a[100];
22
teams_argument_global(int n)23 int teams_argument_global(int n) {
24 int i;
25 int te = n / 128;
26 int th = 128;
27 // discard n_addr and i
28
29
30 #pragma omp target
31 #pragma omp teams distribute simd num_teams(te), thread_limit(th) aligned(a) simdlen(16) linear(i)
32 for(i = 0; i < n; i++) {
33 a[i] = 0;
34 }
35
36 #pragma omp target
37 {{{
38 #pragma omp teams distribute simd safelen(32)
39 for(int i = 0; i < n; i++) {
40 a[i] = 0;
41 }
42 }}}
43
44 // outlined target regions
45
46
47
48
49 return a[0];
50 }
51
52
53 #endif // CK1
54
55 // Test host codegen.
56 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
57 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
58 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
59 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
60 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
61 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
62
63 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
64 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
65 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK13
66 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
67 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
68 // RUN: %clang_cc1 -no-opaque-pointers -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK15
69 #ifdef CK2
70
teams_local_arg(void)71 int teams_local_arg(void) {
72 int n = 100;
73 int a[n];
74
75 #pragma omp target
76 #pragma omp teams distribute simd
77 for(int i = 0; i < n; i++) {
78 a[i] = 0;
79 }
80
81 // outlined target region
82
83
84 return a[0];
85 }
86 #endif // CK2
87
88 // Test host codegen.
89 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
90 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
91 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK17
92 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
93 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
94 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK19
95 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK21
96 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
97 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK21
98 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK23
99 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
100 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK23
101
102 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK25
103 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
104 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK25
105 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK27
106 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
107 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK27
108 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK29
109 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -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
110 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -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 %s --check-prefix=CHECK29
111 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK31
112 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -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
113 // RUN: %clang_cc1 -no-opaque-pointers -DCK3 -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 %s --check-prefix=CHECK31
114 #ifdef CK3
115
116
117 template <typename T, int X, long long Y>
118 struct SS{
119 T a[X];
120 float b;
fooSS121 int foo(void) {
122
123 #pragma omp target
124 #ifdef OMP5
125 #pragma omp teams distribute simd if(b) nontemporal(a, b)
126 #else
127 #pragma omp teams distribute simd
128 #endif // OMP5
129 for(int i = 0; i < X; i++) {
130 a[i] = (T)b;
131 }
132
133 // outlined target region
134
135
136 return a[0];
137 }
138 };
139
teams_template_struct(void)140 int teams_template_struct(void) {
141 SS<int, 123, 456> V;
142 return V.foo();
143
144 }
145 #endif // CK3
146
147 // Test host codegen.
148 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK33
149 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
150 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK33
151 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK35
152 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
153 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK35
154 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK37
155 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
156 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK37
157 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK39
158 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
159 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp -fopenmp-version=50 -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK39
160
161 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK41
162 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
163 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK41
164 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK43
165 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
166 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK43
167 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK45
168 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -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
169 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -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 %s --check-prefix=CHECK45
170 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -verify -fopenmp-simd -fopenmp-version=50 -DOMP5 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK47
171 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -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
172 // RUN: %clang_cc1 -no-opaque-pointers -DCK4 -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 %s --check-prefix=CHECK47
173
174 #ifdef CK4
175
176 template <typename T, int n>
tmain(T argc)177 int tmain(T argc) {
178 T a[n];
179 int te = n/128;
180 int th = 128;
181 #pragma omp target
182 #pragma omp teams distribute simd num_teams(te) thread_limit(th)
183 for(int i = 0; i < n; i++) {
184 a[i] = (T)0;
185 }
186 return 0;
187 }
188
main(int argc,char ** argv)189 int main (int argc, char **argv) {
190 int n = 100;
191 int a[n];
192 #pragma omp target
193 #ifdef OMP5
194 #pragma omp teams distribute simd if(simd:argc)
195 #else
196 #pragma omp teams distribute simd
197 #endif // OMP5
198 for(int i = 0; i < n; i++) {
199 a[i] = 0;
200 }
201 return tmain<int, 10>(argc);
202 }
203
204
205
206
207
208
209
210 // OMP5_50-DAG: !{!"llvm.loop.vectorize.enable", i1 false}
211 #endif // CK4
212 #endif
213
214 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
215 // CHECK1-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
216 // CHECK1-NEXT: entry:
217 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
218 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
219 // CHECK1-NEXT: [[TE:%.*]] = alloca i32, align 4
220 // CHECK1-NEXT: [[TH:%.*]] = alloca i32, align 4
221 // CHECK1-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
222 // CHECK1-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
223 // CHECK1-NEXT: [[I_CASTED:%.*]] = alloca i64, align 8
224 // CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
225 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
226 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
227 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
228 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
229 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
230 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT: [[N_CASTED7:%.*]] = alloca i64, align 8
232 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS9:%.*]] = alloca [2 x i8*], align 8
233 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS10:%.*]] = alloca [2 x i8*], align 8
234 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS11:%.*]] = alloca [2 x i8*], align 8
235 // CHECK1-NEXT: [[_TMP12:%.*]] = alloca i32, align 4
236 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
237 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_14:%.*]] = alloca i32, align 4
238 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
239 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
240 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
241 // CHECK1-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
242 // CHECK1-NEXT: store i32 128, i32* [[TH]], align 4
243 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
244 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
245 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
246 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[TE_CASTED]], align 8
247 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
248 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
249 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[CONV1]], align 4
250 // CHECK1-NEXT: [[TMP4:%.*]] = load i64, i64* [[TH_CASTED]], align 8
251 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[I]], align 4
252 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[I_CASTED]] to i32*
253 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[CONV2]], align 4
254 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[I_CASTED]], align 8
255 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
256 // CHECK1-NEXT: [[CONV3:%.*]] = bitcast i64* [[N_CASTED]] to i32*
257 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[CONV3]], align 4
258 // CHECK1-NEXT: [[TMP8:%.*]] = load i64, i64* [[N_CASTED]], align 8
259 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
260 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
261 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP10]], align 8
262 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
263 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
264 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP12]], align 8
265 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
266 // CHECK1-NEXT: store i8* null, i8** [[TMP13]], align 8
267 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
268 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
269 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP15]], align 8
270 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
271 // CHECK1-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
272 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP17]], align 8
273 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
274 // CHECK1-NEXT: store i8* null, i8** [[TMP18]], align 8
275 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
276 // CHECK1-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [100 x i32]**
277 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP20]], align 8
278 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
279 // CHECK1-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [100 x i32]**
280 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP22]], align 8
281 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
282 // CHECK1-NEXT: store i8* null, i8** [[TMP23]], align 8
283 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
284 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i64*
285 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP25]], align 8
286 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
287 // CHECK1-NEXT: [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i64*
288 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP27]], align 8
289 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
290 // CHECK1-NEXT: store i8* null, i8** [[TMP28]], align 8
291 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
292 // CHECK1-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i64*
293 // CHECK1-NEXT: store i64 [[TMP8]], i64* [[TMP30]], align 8
294 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
295 // CHECK1-NEXT: [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i64*
296 // CHECK1-NEXT: store i64 [[TMP8]], i64* [[TMP32]], align 8
297 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
298 // CHECK1-NEXT: store i8* null, i8** [[TMP33]], align 8
299 // CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
300 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
301 // CHECK1-NEXT: [[TMP36:%.*]] = load i32, i32* [[TE]], align 4
302 // CHECK1-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
303 // CHECK1-NEXT: store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_]], align 4
304 // CHECK1-NEXT: [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
305 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP38]], 0
306 // CHECK1-NEXT: [[DIV5:%.*]] = sdiv i32 [[SUB]], 1
307 // CHECK1-NEXT: [[SUB6:%.*]] = sub nsw i32 [[DIV5]], 1
308 // CHECK1-NEXT: store i32 [[SUB6]], i32* [[DOTCAPTURE_EXPR_4]], align 4
309 // CHECK1-NEXT: [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
310 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP39]], 1
311 // CHECK1-NEXT: [[TMP40:%.*]] = zext i32 [[ADD]] to i64
312 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
313 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
314 // CHECK1-NEXT: store i32 1, i32* [[TMP41]], align 4
315 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
316 // CHECK1-NEXT: store i32 5, i32* [[TMP42]], align 4
317 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
318 // CHECK1-NEXT: store i8** [[TMP34]], i8*** [[TMP43]], align 8
319 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
320 // CHECK1-NEXT: store i8** [[TMP35]], i8*** [[TMP44]], align 8
321 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
322 // CHECK1-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP45]], align 8
323 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
324 // CHECK1-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP46]], align 8
325 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
326 // CHECK1-NEXT: store i8** null, i8*** [[TMP47]], align 8
327 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
328 // CHECK1-NEXT: store i8** null, i8*** [[TMP48]], align 8
329 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
330 // CHECK1-NEXT: store i64 [[TMP40]], i64* [[TMP49]], align 8
331 // CHECK1-NEXT: [[TMP50:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP36]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
332 // CHECK1-NEXT: [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
333 // CHECK1-NEXT: br i1 [[TMP51]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
334 // CHECK1: omp_offload.failed:
335 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30(i64 [[TMP2]], i64 [[TMP4]], [100 x i32]* @a, i64 [[TMP6]], i64 [[TMP8]]) #[[ATTR3:[0-9]+]]
336 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
337 // CHECK1: omp_offload.cont:
338 // CHECK1-NEXT: [[TMP52:%.*]] = load i32, i32* [[N_ADDR]], align 4
339 // CHECK1-NEXT: [[CONV8:%.*]] = bitcast i64* [[N_CASTED7]] to i32*
340 // CHECK1-NEXT: store i32 [[TMP52]], i32* [[CONV8]], align 4
341 // CHECK1-NEXT: [[TMP53:%.*]] = load i64, i64* [[N_CASTED7]], align 8
342 // CHECK1-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
343 // CHECK1-NEXT: [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i64*
344 // CHECK1-NEXT: store i64 [[TMP53]], i64* [[TMP55]], align 8
345 // CHECK1-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
346 // CHECK1-NEXT: [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i64*
347 // CHECK1-NEXT: store i64 [[TMP53]], i64* [[TMP57]], align 8
348 // CHECK1-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 0
349 // CHECK1-NEXT: store i8* null, i8** [[TMP58]], align 8
350 // CHECK1-NEXT: [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 1
351 // CHECK1-NEXT: [[TMP60:%.*]] = bitcast i8** [[TMP59]] to [100 x i32]**
352 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP60]], align 8
353 // CHECK1-NEXT: [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 1
354 // CHECK1-NEXT: [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [100 x i32]**
355 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP62]], align 8
356 // CHECK1-NEXT: [[TMP63:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS11]], i64 0, i64 1
357 // CHECK1-NEXT: store i8* null, i8** [[TMP63]], align 8
358 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS9]], i32 0, i32 0
359 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS10]], i32 0, i32 0
360 // CHECK1-NEXT: [[TMP66:%.*]] = load i32, i32* [[N_ADDR]], align 4
361 // CHECK1-NEXT: store i32 [[TMP66]], i32* [[DOTCAPTURE_EXPR_13]], align 4
362 // CHECK1-NEXT: [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
363 // CHECK1-NEXT: [[SUB15:%.*]] = sub nsw i32 [[TMP67]], 0
364 // CHECK1-NEXT: [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
365 // CHECK1-NEXT: [[SUB17:%.*]] = sub nsw i32 [[DIV16]], 1
366 // CHECK1-NEXT: store i32 [[SUB17]], i32* [[DOTCAPTURE_EXPR_14]], align 4
367 // CHECK1-NEXT: [[TMP68:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
368 // CHECK1-NEXT: [[ADD18:%.*]] = add nsw i32 [[TMP68]], 1
369 // CHECK1-NEXT: [[TMP69:%.*]] = zext i32 [[ADD18]] to i64
370 // CHECK1-NEXT: [[KERNEL_ARGS19:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
371 // CHECK1-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 0
372 // CHECK1-NEXT: store i32 1, i32* [[TMP70]], align 4
373 // CHECK1-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 1
374 // CHECK1-NEXT: store i32 2, i32* [[TMP71]], align 4
375 // CHECK1-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 2
376 // CHECK1-NEXT: store i8** [[TMP64]], i8*** [[TMP72]], align 8
377 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 3
378 // CHECK1-NEXT: store i8** [[TMP65]], i8*** [[TMP73]], align 8
379 // CHECK1-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 4
380 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP74]], align 8
381 // CHECK1-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 5
382 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP75]], align 8
383 // CHECK1-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 6
384 // CHECK1-NEXT: store i8** null, i8*** [[TMP76]], align 8
385 // CHECK1-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 7
386 // CHECK1-NEXT: store i8** null, i8*** [[TMP77]], align 8
387 // CHECK1-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]], i32 0, i32 8
388 // CHECK1-NEXT: store i64 [[TMP69]], i64* [[TMP78]], align 8
389 // CHECK1-NEXT: [[TMP79:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS19]])
390 // CHECK1-NEXT: [[TMP80:%.*]] = icmp ne i32 [[TMP79]], 0
391 // CHECK1-NEXT: br i1 [[TMP80]], label [[OMP_OFFLOAD_FAILED20:%.*]], label [[OMP_OFFLOAD_CONT21:%.*]]
392 // CHECK1: omp_offload.failed20:
393 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36(i64 [[TMP53]], [100 x i32]* @a) #[[ATTR3]]
394 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT21]]
395 // CHECK1: omp_offload.cont21:
396 // CHECK1-NEXT: [[TMP81:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4
397 // CHECK1-NEXT: ret i32 [[TMP81]]
398 //
399 //
400 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30
401 // CHECK1-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]], i64 noundef [[I:%.*]], i64 noundef [[N:%.*]]) #[[ATTR1:[0-9]+]] {
402 // CHECK1-NEXT: entry:
403 // CHECK1-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
404 // CHECK1-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
405 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
406 // CHECK1-NEXT: [[I_ADDR:%.*]] = alloca i64, align 8
407 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
408 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
409 // CHECK1-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
410 // CHECK1-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
411 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
412 // CHECK1-NEXT: store i64 [[I]], i64* [[I_ADDR]], align 8
413 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
414 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
415 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
416 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
417 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[I_ADDR]] to i32*
418 // CHECK1-NEXT: [[CONV3:%.*]] = bitcast i64* [[N_ADDR]] to i32*
419 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
420 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
421 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
422 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV2]], i32* [[CONV3]], [100 x i32]* [[TMP1]])
423 // CHECK1-NEXT: ret void
424 //
425 //
426 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
427 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
428 // CHECK1-NEXT: entry:
429 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
430 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
431 // CHECK1-NEXT: [[I_ADDR:%.*]] = alloca i32*, align 8
432 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
433 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
434 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
435 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
436 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
437 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
438 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
439 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
440 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
441 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
442 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
443 // CHECK1-NEXT: [[I4:%.*]] = alloca i32, align 4
444 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
445 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
446 // CHECK1-NEXT: store i32* [[I]], i32** [[I_ADDR]], align 8
447 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
448 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
449 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 8
450 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[N_ADDR]], align 8
451 // CHECK1-NEXT: [[TMP2:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
452 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
453 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
454 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
455 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
456 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
457 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
458 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
459 // CHECK1-NEXT: store i32 0, i32* [[I3]], align 4
460 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
461 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
462 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
463 // CHECK1: omp.precond.then:
464 // CHECK1-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i64 0, i64 0
465 // CHECK1-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAYDECAY]], i64 16) ]
466 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
467 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
468 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
469 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
470 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
471 // CHECK1-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
472 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
473 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
474 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
475 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
476 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
477 // CHECK1-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
478 // CHECK1: cond.true:
479 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
480 // CHECK1-NEXT: br label [[COND_END:%.*]]
481 // CHECK1: cond.false:
482 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
483 // CHECK1-NEXT: br label [[COND_END]]
484 // CHECK1: cond.end:
485 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
486 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
487 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
488 // CHECK1-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
489 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
490 // CHECK1: omp.inner.for.cond:
491 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
492 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
493 // CHECK1-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
494 // CHECK1-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
495 // CHECK1: omp.inner.for.body:
496 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
497 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
498 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
499 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !5
500 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !5
501 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
502 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i64 0, i64 [[IDXPROM]]
503 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
504 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
505 // CHECK1: omp.body.continue:
506 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
507 // CHECK1: omp.inner.for.inc:
508 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
509 // CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP18]], 1
510 // CHECK1-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
511 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
512 // CHECK1: omp.inner.for.end:
513 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
514 // CHECK1: omp.loop.exit:
515 // CHECK1-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
516 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
517 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
518 // CHECK1-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
519 // CHECK1-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
520 // CHECK1-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
521 // CHECK1: .omp.final.then:
522 // CHECK1-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
523 // CHECK1-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP23]], 0
524 // CHECK1-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
525 // CHECK1-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
526 // CHECK1-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
527 // CHECK1-NEXT: store i32 [[ADD11]], i32* [[TMP0]], align 4
528 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
529 // CHECK1: .omp.final.done:
530 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
531 // CHECK1: omp.precond.end:
532 // CHECK1-NEXT: ret void
533 //
534 //
535 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36
536 // CHECK1-SAME: (i64 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
537 // CHECK1-NEXT: entry:
538 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
539 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
540 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
541 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
542 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
543 // CHECK1-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
544 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[CONV]], [100 x i32]* [[TMP0]])
545 // CHECK1-NEXT: ret void
546 //
547 //
548 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
549 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
550 // CHECK1-NEXT: entry:
551 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
552 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
553 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
554 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
555 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
556 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
557 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
558 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
559 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
560 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
561 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
562 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
563 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
564 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
565 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
566 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
567 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
568 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
569 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
570 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
571 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
572 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
573 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
574 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
575 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
576 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
577 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
578 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4
579 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
580 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
581 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
582 // CHECK1: omp.precond.then:
583 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
584 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
585 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
586 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
587 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
588 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
589 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
590 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
591 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
592 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
593 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
594 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
595 // CHECK1: cond.true:
596 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
597 // CHECK1-NEXT: br label [[COND_END:%.*]]
598 // CHECK1: cond.false:
599 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
600 // CHECK1-NEXT: br label [[COND_END]]
601 // CHECK1: cond.end:
602 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
603 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
604 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
605 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
606 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
607 // CHECK1: omp.inner.for.cond:
608 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
609 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
610 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
611 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
612 // CHECK1: omp.inner.for.body:
613 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
614 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
615 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
616 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
617 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
618 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
619 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
620 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
621 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
622 // CHECK1: omp.body.continue:
623 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
624 // CHECK1: omp.inner.for.inc:
625 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
626 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
627 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
628 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
629 // CHECK1: omp.inner.for.end:
630 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
631 // CHECK1: omp.loop.exit:
632 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
633 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
634 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
635 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
636 // CHECK1-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
637 // CHECK1-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
638 // CHECK1: .omp.final.then:
639 // CHECK1-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
640 // CHECK1-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP22]], 0
641 // CHECK1-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
642 // CHECK1-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
643 // CHECK1-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
644 // CHECK1-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
645 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
646 // CHECK1: .omp.final.done:
647 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
648 // CHECK1: omp.precond.end:
649 // CHECK1-NEXT: ret void
650 //
651 //
652 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
653 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
654 // CHECK1-NEXT: entry:
655 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
656 // CHECK1-NEXT: ret void
657 //
658 //
659 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
660 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
661 // CHECK3-NEXT: entry:
662 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
663 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
664 // CHECK3-NEXT: [[TE:%.*]] = alloca i32, align 4
665 // CHECK3-NEXT: [[TH:%.*]] = alloca i32, align 4
666 // CHECK3-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
667 // CHECK3-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
668 // CHECK3-NEXT: [[I_CASTED:%.*]] = alloca i32, align 4
669 // CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
670 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
671 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
672 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
673 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
674 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
675 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
676 // CHECK3-NEXT: [[N_CASTED4:%.*]] = alloca i32, align 4
677 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
678 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
679 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
680 // CHECK3-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
681 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
682 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
683 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
684 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
685 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
686 // CHECK3-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
687 // CHECK3-NEXT: store i32 128, i32* [[TH]], align 4
688 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
689 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TE_CASTED]], align 4
690 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_CASTED]], align 4
691 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
692 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TH_CASTED]], align 4
693 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TH_CASTED]], align 4
694 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[I]], align 4
695 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[I_CASTED]], align 4
696 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[I_CASTED]], align 4
697 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[N_ADDR]], align 4
698 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[N_CASTED]], align 4
699 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[N_CASTED]], align 4
700 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
701 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
702 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP10]], align 4
703 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
704 // CHECK3-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
705 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP12]], align 4
706 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
707 // CHECK3-NEXT: store i8* null, i8** [[TMP13]], align 4
708 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
709 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
710 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP15]], align 4
711 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
712 // CHECK3-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
713 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP17]], align 4
714 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
715 // CHECK3-NEXT: store i8* null, i8** [[TMP18]], align 4
716 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
717 // CHECK3-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to [100 x i32]**
718 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP20]], align 4
719 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
720 // CHECK3-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [100 x i32]**
721 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP22]], align 4
722 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
723 // CHECK3-NEXT: store i8* null, i8** [[TMP23]], align 4
724 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
725 // CHECK3-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
726 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP25]], align 4
727 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
728 // CHECK3-NEXT: [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32*
729 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP27]], align 4
730 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
731 // CHECK3-NEXT: store i8* null, i8** [[TMP28]], align 4
732 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
733 // CHECK3-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
734 // CHECK3-NEXT: store i32 [[TMP8]], i32* [[TMP30]], align 4
735 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
736 // CHECK3-NEXT: [[TMP32:%.*]] = bitcast i8** [[TMP31]] to i32*
737 // CHECK3-NEXT: store i32 [[TMP8]], i32* [[TMP32]], align 4
738 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
739 // CHECK3-NEXT: store i8* null, i8** [[TMP33]], align 4
740 // CHECK3-NEXT: [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
741 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
742 // CHECK3-NEXT: [[TMP36:%.*]] = load i32, i32* [[TE]], align 4
743 // CHECK3-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
744 // CHECK3-NEXT: store i32 [[TMP37]], i32* [[DOTCAPTURE_EXPR_]], align 4
745 // CHECK3-NEXT: [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
746 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP38]], 0
747 // CHECK3-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
748 // CHECK3-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
749 // CHECK3-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
750 // CHECK3-NEXT: [[TMP39:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
751 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP39]], 1
752 // CHECK3-NEXT: [[TMP40:%.*]] = zext i32 [[ADD]] to i64
753 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
754 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
755 // CHECK3-NEXT: store i32 1, i32* [[TMP41]], align 4
756 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
757 // CHECK3-NEXT: store i32 5, i32* [[TMP42]], align 4
758 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
759 // CHECK3-NEXT: store i8** [[TMP34]], i8*** [[TMP43]], align 4
760 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
761 // CHECK3-NEXT: store i8** [[TMP35]], i8*** [[TMP44]], align 4
762 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
763 // CHECK3-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP45]], align 4
764 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
765 // CHECK3-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP46]], align 4
766 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
767 // CHECK3-NEXT: store i8** null, i8*** [[TMP47]], align 4
768 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
769 // CHECK3-NEXT: store i8** null, i8*** [[TMP48]], align 4
770 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
771 // CHECK3-NEXT: store i64 [[TMP40]], i64* [[TMP49]], align 8
772 // CHECK3-NEXT: [[TMP50:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP36]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
773 // CHECK3-NEXT: [[TMP51:%.*]] = icmp ne i32 [[TMP50]], 0
774 // CHECK3-NEXT: br i1 [[TMP51]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
775 // CHECK3: omp_offload.failed:
776 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30(i32 [[TMP2]], i32 [[TMP4]], [100 x i32]* @a, i32 [[TMP6]], i32 [[TMP8]]) #[[ATTR3:[0-9]+]]
777 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
778 // CHECK3: omp_offload.cont:
779 // CHECK3-NEXT: [[TMP52:%.*]] = load i32, i32* [[N_ADDR]], align 4
780 // CHECK3-NEXT: store i32 [[TMP52]], i32* [[N_CASTED4]], align 4
781 // CHECK3-NEXT: [[TMP53:%.*]] = load i32, i32* [[N_CASTED4]], align 4
782 // CHECK3-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
783 // CHECK3-NEXT: [[TMP55:%.*]] = bitcast i8** [[TMP54]] to i32*
784 // CHECK3-NEXT: store i32 [[TMP53]], i32* [[TMP55]], align 4
785 // CHECK3-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
786 // CHECK3-NEXT: [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32*
787 // CHECK3-NEXT: store i32 [[TMP53]], i32* [[TMP57]], align 4
788 // CHECK3-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
789 // CHECK3-NEXT: store i8* null, i8** [[TMP58]], align 4
790 // CHECK3-NEXT: [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
791 // CHECK3-NEXT: [[TMP60:%.*]] = bitcast i8** [[TMP59]] to [100 x i32]**
792 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP60]], align 4
793 // CHECK3-NEXT: [[TMP61:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
794 // CHECK3-NEXT: [[TMP62:%.*]] = bitcast i8** [[TMP61]] to [100 x i32]**
795 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP62]], align 4
796 // CHECK3-NEXT: [[TMP63:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
797 // CHECK3-NEXT: store i8* null, i8** [[TMP63]], align 4
798 // CHECK3-NEXT: [[TMP64:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
799 // CHECK3-NEXT: [[TMP65:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
800 // CHECK3-NEXT: [[TMP66:%.*]] = load i32, i32* [[N_ADDR]], align 4
801 // CHECK3-NEXT: store i32 [[TMP66]], i32* [[DOTCAPTURE_EXPR_9]], align 4
802 // CHECK3-NEXT: [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
803 // CHECK3-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP67]], 0
804 // CHECK3-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
805 // CHECK3-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
806 // CHECK3-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
807 // CHECK3-NEXT: [[TMP68:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
808 // CHECK3-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP68]], 1
809 // CHECK3-NEXT: [[TMP69:%.*]] = zext i32 [[ADD14]] to i64
810 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
811 // CHECK3-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
812 // CHECK3-NEXT: store i32 1, i32* [[TMP70]], align 4
813 // CHECK3-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
814 // CHECK3-NEXT: store i32 2, i32* [[TMP71]], align 4
815 // CHECK3-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
816 // CHECK3-NEXT: store i8** [[TMP64]], i8*** [[TMP72]], align 4
817 // CHECK3-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
818 // CHECK3-NEXT: store i8** [[TMP65]], i8*** [[TMP73]], align 4
819 // CHECK3-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
820 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP74]], align 4
821 // CHECK3-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
822 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP75]], align 4
823 // CHECK3-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
824 // CHECK3-NEXT: store i8** null, i8*** [[TMP76]], align 4
825 // CHECK3-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
826 // CHECK3-NEXT: store i8** null, i8*** [[TMP77]], align 4
827 // CHECK3-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
828 // CHECK3-NEXT: store i64 [[TMP69]], i64* [[TMP78]], align 8
829 // CHECK3-NEXT: [[TMP79:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
830 // CHECK3-NEXT: [[TMP80:%.*]] = icmp ne i32 [[TMP79]], 0
831 // CHECK3-NEXT: br i1 [[TMP80]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
832 // CHECK3: omp_offload.failed16:
833 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36(i32 [[TMP53]], [100 x i32]* @a) #[[ATTR3]]
834 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]]
835 // CHECK3: omp_offload.cont17:
836 // CHECK3-NEXT: [[TMP81:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4
837 // CHECK3-NEXT: ret i32 [[TMP81]]
838 //
839 //
840 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l30
841 // CHECK3-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]], i32 noundef [[I:%.*]], i32 noundef [[N:%.*]]) #[[ATTR1:[0-9]+]] {
842 // CHECK3-NEXT: entry:
843 // CHECK3-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
844 // CHECK3-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
845 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
846 // CHECK3-NEXT: [[I_ADDR:%.*]] = alloca i32, align 4
847 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
848 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
849 // CHECK3-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
850 // CHECK3-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
851 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
852 // CHECK3-NEXT: store i32 [[I]], i32* [[I_ADDR]], align 4
853 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
854 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
855 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
856 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
857 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
858 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[I_ADDR]], i32* [[N_ADDR]], [100 x i32]* [[TMP1]])
859 // CHECK3-NEXT: ret void
860 //
861 //
862 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
863 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[I:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
864 // CHECK3-NEXT: entry:
865 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
866 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
867 // CHECK3-NEXT: [[I_ADDR:%.*]] = alloca i32*, align 4
868 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
869 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
870 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
871 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
872 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
873 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
874 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
875 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
876 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
877 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
878 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
879 // CHECK3-NEXT: [[I4:%.*]] = alloca i32, align 4
880 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
881 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
882 // CHECK3-NEXT: store i32* [[I]], i32** [[I_ADDR]], align 4
883 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
884 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
885 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[I_ADDR]], align 4
886 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[N_ADDR]], align 4
887 // CHECK3-NEXT: [[TMP2:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
888 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP1]], align 4
889 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
890 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
891 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
892 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
893 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
894 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
895 // CHECK3-NEXT: store i32 0, i32* [[I3]], align 4
896 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
897 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
898 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
899 // CHECK3: omp.precond.then:
900 // CHECK3-NEXT: [[ARRAYDECAY:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i32 0, i32 0
901 // CHECK3-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* [[ARRAYDECAY]], i32 16) ]
902 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
903 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
904 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
905 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
906 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
907 // CHECK3-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
908 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
909 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
910 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
911 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
912 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
913 // CHECK3-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
914 // CHECK3: cond.true:
915 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
916 // CHECK3-NEXT: br label [[COND_END:%.*]]
917 // CHECK3: cond.false:
918 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
919 // CHECK3-NEXT: br label [[COND_END]]
920 // CHECK3: cond.end:
921 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
922 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
923 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
924 // CHECK3-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
925 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
926 // CHECK3: omp.inner.for.cond:
927 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
928 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
929 // CHECK3-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
930 // CHECK3-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
931 // CHECK3: omp.inner.for.body:
932 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
933 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
934 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
935 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !6
936 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !6
937 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP2]], i32 0, i32 [[TMP17]]
938 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6
939 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
940 // CHECK3: omp.body.continue:
941 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
942 // CHECK3: omp.inner.for.inc:
943 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
944 // CHECK3-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP18]], 1
945 // CHECK3-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
946 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
947 // CHECK3: omp.inner.for.end:
948 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
949 // CHECK3: omp.loop.exit:
950 // CHECK3-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
951 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
952 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
953 // CHECK3-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
954 // CHECK3-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
955 // CHECK3-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
956 // CHECK3: .omp.final.then:
957 // CHECK3-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
958 // CHECK3-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP23]], 0
959 // CHECK3-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
960 // CHECK3-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
961 // CHECK3-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
962 // CHECK3-NEXT: store i32 [[ADD11]], i32* [[TMP0]], align 4
963 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
964 // CHECK3: .omp.final.done:
965 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
966 // CHECK3: omp.precond.end:
967 // CHECK3-NEXT: ret void
968 //
969 //
970 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l36
971 // CHECK3-SAME: (i32 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
972 // CHECK3-NEXT: entry:
973 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
974 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
975 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
976 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
977 // CHECK3-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
978 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP0]])
979 // CHECK3-NEXT: ret void
980 //
981 //
982 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
983 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
984 // CHECK3-NEXT: entry:
985 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
986 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
987 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
988 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
989 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
990 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
991 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
992 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
993 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
994 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
995 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
996 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
997 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
998 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
999 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1000 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1001 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1002 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1003 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1004 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1005 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1006 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1007 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1008 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1009 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1010 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1011 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1012 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4
1013 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1014 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1015 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1016 // CHECK3: omp.precond.then:
1017 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1018 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1019 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1020 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1021 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1022 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1023 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1024 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1025 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1026 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1027 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1028 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1029 // CHECK3: cond.true:
1030 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1031 // CHECK3-NEXT: br label [[COND_END:%.*]]
1032 // CHECK3: cond.false:
1033 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1034 // CHECK3-NEXT: br label [[COND_END]]
1035 // CHECK3: cond.end:
1036 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1037 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1038 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1039 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1040 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1041 // CHECK3: omp.inner.for.cond:
1042 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1043 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1044 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1045 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1046 // CHECK3: omp.inner.for.body:
1047 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1048 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1049 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1050 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1051 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
1052 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
1053 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1054 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1055 // CHECK3: omp.body.continue:
1056 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1057 // CHECK3: omp.inner.for.inc:
1058 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1059 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1060 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1061 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1062 // CHECK3: omp.inner.for.end:
1063 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1064 // CHECK3: omp.loop.exit:
1065 // CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1066 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
1067 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
1068 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1069 // CHECK3-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
1070 // CHECK3-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1071 // CHECK3: .omp.final.then:
1072 // CHECK3-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1073 // CHECK3-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP22]], 0
1074 // CHECK3-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
1075 // CHECK3-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
1076 // CHECK3-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
1077 // CHECK3-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
1078 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1079 // CHECK3: .omp.final.done:
1080 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
1081 // CHECK3: omp.precond.end:
1082 // CHECK3-NEXT: ret void
1083 //
1084 //
1085 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1086 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
1087 // CHECK3-NEXT: entry:
1088 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1089 // CHECK3-NEXT: ret void
1090 //
1091 //
1092 // CHECK5-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
1093 // CHECK5-SAME: (i32 noundef signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1094 // CHECK5-NEXT: entry:
1095 // CHECK5-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1096 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
1097 // CHECK5-NEXT: [[TE:%.*]] = alloca i32, align 4
1098 // CHECK5-NEXT: [[TH:%.*]] = alloca i32, align 4
1099 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
1100 // CHECK5-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1101 // CHECK5-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1102 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1103 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1104 // CHECK5-NEXT: [[I4:%.*]] = alloca i32, align 4
1105 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1106 // CHECK5-NEXT: [[DOTLINEAR_START:%.*]] = alloca i32, align 4
1107 // CHECK5-NEXT: [[I5:%.*]] = alloca i32, align 4
1108 // CHECK5-NEXT: [[I6:%.*]] = alloca i32, align 4
1109 // CHECK5-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
1110 // CHECK5-NEXT: [[DOTCAPTURE_EXPR_14:%.*]] = alloca i32, align 4
1111 // CHECK5-NEXT: [[DOTCAPTURE_EXPR_15:%.*]] = alloca i32, align 4
1112 // CHECK5-NEXT: [[DOTOMP_LB19:%.*]] = alloca i32, align 4
1113 // CHECK5-NEXT: [[DOTOMP_UB20:%.*]] = alloca i32, align 4
1114 // CHECK5-NEXT: [[I21:%.*]] = alloca i32, align 4
1115 // CHECK5-NEXT: [[DOTOMP_IV24:%.*]] = alloca i32, align 4
1116 // CHECK5-NEXT: [[I25:%.*]] = alloca i32, align 4
1117 // CHECK5-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1118 // CHECK5-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1119 // CHECK5-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
1120 // CHECK5-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
1121 // CHECK5-NEXT: store i32 128, i32* [[TH]], align 4
1122 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1123 // CHECK5-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
1124 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1125 // CHECK5-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
1126 // CHECK5-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
1127 // CHECK5-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
1128 // CHECK5-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1129 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1130 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1131 // CHECK5-NEXT: store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
1132 // CHECK5-NEXT: store i32 0, i32* [[I4]], align 4
1133 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1134 // CHECK5-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1135 // CHECK5-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
1136 // CHECK5: simd.if.then:
1137 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1138 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1139 // CHECK5-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), i64 16) ]
1140 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4
1141 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTLINEAR_START]], align 4
1142 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1143 // CHECK5: omp.inner.for.cond:
1144 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1145 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1146 // CHECK5-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1147 // CHECK5-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1148 // CHECK5: omp.inner.for.body:
1149 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1150 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1151 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1152 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I5]], align 4, !llvm.access.group !2
1153 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !2
1154 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP10]] to i64
1155 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* @a, i64 0, i64 [[IDXPROM]]
1156 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
1157 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1158 // CHECK5: omp.body.continue:
1159 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1160 // CHECK5: omp.inner.for.inc:
1161 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1162 // CHECK5-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP11]], 1
1163 // CHECK5-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1164 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1165 // CHECK5: omp.inner.for.end:
1166 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1167 // CHECK5-NEXT: [[SUB9:%.*]] = sub nsw i32 [[TMP12]], 0
1168 // CHECK5-NEXT: [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
1169 // CHECK5-NEXT: [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
1170 // CHECK5-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
1171 // CHECK5-NEXT: store i32 [[ADD12]], i32* [[I]], align 4
1172 // CHECK5-NEXT: br label [[SIMD_IF_END]]
1173 // CHECK5: simd.if.end:
1174 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
1175 // CHECK5-NEXT: store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_14]], align 4
1176 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1177 // CHECK5-NEXT: [[SUB16:%.*]] = sub nsw i32 [[TMP14]], 0
1178 // CHECK5-NEXT: [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
1179 // CHECK5-NEXT: [[SUB18:%.*]] = sub nsw i32 [[DIV17]], 1
1180 // CHECK5-NEXT: store i32 [[SUB18]], i32* [[DOTCAPTURE_EXPR_15]], align 4
1181 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB19]], align 4
1182 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_15]], align 4
1183 // CHECK5-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_UB20]], align 4
1184 // CHECK5-NEXT: store i32 0, i32* [[I21]], align 4
1185 // CHECK5-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1186 // CHECK5-NEXT: [[CMP22:%.*]] = icmp slt i32 0, [[TMP16]]
1187 // CHECK5-NEXT: br i1 [[CMP22]], label [[SIMD_IF_THEN23:%.*]], label [[SIMD_IF_END41:%.*]]
1188 // CHECK5: simd.if.then23:
1189 // CHECK5-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
1190 // CHECK5-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV24]], align 4
1191 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND26:%.*]]
1192 // CHECK5: omp.inner.for.cond26:
1193 // CHECK5-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1194 // CHECK5-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4
1195 // CHECK5-NEXT: [[CMP27:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
1196 // CHECK5-NEXT: br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
1197 // CHECK5: omp.inner.for.body28:
1198 // CHECK5-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1199 // CHECK5-NEXT: [[MUL29:%.*]] = mul nsw i32 [[TMP20]], 1
1200 // CHECK5-NEXT: [[ADD30:%.*]] = add nsw i32 0, [[MUL29]]
1201 // CHECK5-NEXT: store i32 [[ADD30]], i32* [[I25]], align 4
1202 // CHECK5-NEXT: [[TMP21:%.*]] = load i32, i32* [[I25]], align 4
1203 // CHECK5-NEXT: [[IDXPROM31:%.*]] = sext i32 [[TMP21]] to i64
1204 // CHECK5-NEXT: [[ARRAYIDX32:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* @a, i64 0, i64 [[IDXPROM31]]
1205 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX32]], align 4
1206 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE33:%.*]]
1207 // CHECK5: omp.body.continue33:
1208 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC34:%.*]]
1209 // CHECK5: omp.inner.for.inc34:
1210 // CHECK5-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1211 // CHECK5-NEXT: [[ADD35:%.*]] = add nsw i32 [[TMP22]], 1
1212 // CHECK5-NEXT: store i32 [[ADD35]], i32* [[DOTOMP_IV24]], align 4
1213 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND26]], !llvm.loop [[LOOP7:![0-9]+]]
1214 // CHECK5: omp.inner.for.end36:
1215 // CHECK5-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1216 // CHECK5-NEXT: [[SUB37:%.*]] = sub nsw i32 [[TMP23]], 0
1217 // CHECK5-NEXT: [[DIV38:%.*]] = sdiv i32 [[SUB37]], 1
1218 // CHECK5-NEXT: [[MUL39:%.*]] = mul nsw i32 [[DIV38]], 1
1219 // CHECK5-NEXT: [[ADD40:%.*]] = add nsw i32 0, [[MUL39]]
1220 // CHECK5-NEXT: store i32 [[ADD40]], i32* [[I25]], align 4
1221 // CHECK5-NEXT: br label [[SIMD_IF_END41]]
1222 // CHECK5: simd.if.end41:
1223 // CHECK5-NEXT: [[TMP24:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4
1224 // CHECK5-NEXT: ret i32 [[TMP24]]
1225 //
1226 //
1227 // CHECK7-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
1228 // CHECK7-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1229 // CHECK7-NEXT: entry:
1230 // CHECK7-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1231 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
1232 // CHECK7-NEXT: [[TE:%.*]] = alloca i32, align 4
1233 // CHECK7-NEXT: [[TH:%.*]] = alloca i32, align 4
1234 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
1235 // CHECK7-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1236 // CHECK7-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1237 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1238 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1239 // CHECK7-NEXT: [[I4:%.*]] = alloca i32, align 4
1240 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1241 // CHECK7-NEXT: [[DOTLINEAR_START:%.*]] = alloca i32, align 4
1242 // CHECK7-NEXT: [[I5:%.*]] = alloca i32, align 4
1243 // CHECK7-NEXT: [[I6:%.*]] = alloca i32, align 4
1244 // CHECK7-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
1245 // CHECK7-NEXT: [[DOTCAPTURE_EXPR_14:%.*]] = alloca i32, align 4
1246 // CHECK7-NEXT: [[DOTCAPTURE_EXPR_15:%.*]] = alloca i32, align 4
1247 // CHECK7-NEXT: [[DOTOMP_LB19:%.*]] = alloca i32, align 4
1248 // CHECK7-NEXT: [[DOTOMP_UB20:%.*]] = alloca i32, align 4
1249 // CHECK7-NEXT: [[I21:%.*]] = alloca i32, align 4
1250 // CHECK7-NEXT: [[DOTOMP_IV24:%.*]] = alloca i32, align 4
1251 // CHECK7-NEXT: [[I25:%.*]] = alloca i32, align 4
1252 // CHECK7-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1253 // CHECK7-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1254 // CHECK7-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
1255 // CHECK7-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
1256 // CHECK7-NEXT: store i32 128, i32* [[TH]], align 4
1257 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[N_ADDR]], align 4
1258 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
1259 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1260 // CHECK7-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP2]], 0
1261 // CHECK7-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
1262 // CHECK7-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
1263 // CHECK7-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1264 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1265 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1266 // CHECK7-NEXT: store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
1267 // CHECK7-NEXT: store i32 0, i32* [[I4]], align 4
1268 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1269 // CHECK7-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1270 // CHECK7-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
1271 // CHECK7: simd.if.then:
1272 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1273 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1274 // CHECK7-NEXT: call void @llvm.assume(i1 true) [ "align"(i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), i32 16) ]
1275 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4
1276 // CHECK7-NEXT: store i32 [[TMP6]], i32* [[DOTLINEAR_START]], align 4
1277 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1278 // CHECK7: omp.inner.for.cond:
1279 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1280 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
1281 // CHECK7-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1282 // CHECK7-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1283 // CHECK7: omp.inner.for.body:
1284 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1285 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1286 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1287 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I5]], align 4, !llvm.access.group !3
1288 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[I5]], align 4, !llvm.access.group !3
1289 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* @a, i32 0, i32 [[TMP10]]
1290 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
1291 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1292 // CHECK7: omp.body.continue:
1293 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1294 // CHECK7: omp.inner.for.inc:
1295 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1296 // CHECK7-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP11]], 1
1297 // CHECK7-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1298 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
1299 // CHECK7: omp.inner.for.end:
1300 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1301 // CHECK7-NEXT: [[SUB9:%.*]] = sub nsw i32 [[TMP12]], 0
1302 // CHECK7-NEXT: [[DIV10:%.*]] = sdiv i32 [[SUB9]], 1
1303 // CHECK7-NEXT: [[MUL11:%.*]] = mul nsw i32 [[DIV10]], 1
1304 // CHECK7-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
1305 // CHECK7-NEXT: store i32 [[ADD12]], i32* [[I]], align 4
1306 // CHECK7-NEXT: br label [[SIMD_IF_END]]
1307 // CHECK7: simd.if.end:
1308 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[N_ADDR]], align 4
1309 // CHECK7-NEXT: store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_14]], align 4
1310 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1311 // CHECK7-NEXT: [[SUB16:%.*]] = sub nsw i32 [[TMP14]], 0
1312 // CHECK7-NEXT: [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
1313 // CHECK7-NEXT: [[SUB18:%.*]] = sub nsw i32 [[DIV17]], 1
1314 // CHECK7-NEXT: store i32 [[SUB18]], i32* [[DOTCAPTURE_EXPR_15]], align 4
1315 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB19]], align 4
1316 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_15]], align 4
1317 // CHECK7-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_UB20]], align 4
1318 // CHECK7-NEXT: store i32 0, i32* [[I21]], align 4
1319 // CHECK7-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1320 // CHECK7-NEXT: [[CMP22:%.*]] = icmp slt i32 0, [[TMP16]]
1321 // CHECK7-NEXT: br i1 [[CMP22]], label [[SIMD_IF_THEN23:%.*]], label [[SIMD_IF_END40:%.*]]
1322 // CHECK7: simd.if.then23:
1323 // CHECK7-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
1324 // CHECK7-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV24]], align 4
1325 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND26:%.*]]
1326 // CHECK7: omp.inner.for.cond26:
1327 // CHECK7-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1328 // CHECK7-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4
1329 // CHECK7-NEXT: [[CMP27:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
1330 // CHECK7-NEXT: br i1 [[CMP27]], label [[OMP_INNER_FOR_BODY28:%.*]], label [[OMP_INNER_FOR_END35:%.*]]
1331 // CHECK7: omp.inner.for.body28:
1332 // CHECK7-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1333 // CHECK7-NEXT: [[MUL29:%.*]] = mul nsw i32 [[TMP20]], 1
1334 // CHECK7-NEXT: [[ADD30:%.*]] = add nsw i32 0, [[MUL29]]
1335 // CHECK7-NEXT: store i32 [[ADD30]], i32* [[I25]], align 4
1336 // CHECK7-NEXT: [[TMP21:%.*]] = load i32, i32* [[I25]], align 4
1337 // CHECK7-NEXT: [[ARRAYIDX31:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* @a, i32 0, i32 [[TMP21]]
1338 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX31]], align 4
1339 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE32:%.*]]
1340 // CHECK7: omp.body.continue32:
1341 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC33:%.*]]
1342 // CHECK7: omp.inner.for.inc33:
1343 // CHECK7-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV24]], align 4
1344 // CHECK7-NEXT: [[ADD34:%.*]] = add nsw i32 [[TMP22]], 1
1345 // CHECK7-NEXT: store i32 [[ADD34]], i32* [[DOTOMP_IV24]], align 4
1346 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND26]], !llvm.loop [[LOOP8:![0-9]+]]
1347 // CHECK7: omp.inner.for.end35:
1348 // CHECK7-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_14]], align 4
1349 // CHECK7-NEXT: [[SUB36:%.*]] = sub nsw i32 [[TMP23]], 0
1350 // CHECK7-NEXT: [[DIV37:%.*]] = sdiv i32 [[SUB36]], 1
1351 // CHECK7-NEXT: [[MUL38:%.*]] = mul nsw i32 [[DIV37]], 1
1352 // CHECK7-NEXT: [[ADD39:%.*]] = add nsw i32 0, [[MUL38]]
1353 // CHECK7-NEXT: store i32 [[ADD39]], i32* [[I25]], align 4
1354 // CHECK7-NEXT: br label [[SIMD_IF_END40]]
1355 // CHECK7: simd.if.end40:
1356 // CHECK7-NEXT: [[TMP24:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4
1357 // CHECK7-NEXT: ret i32 [[TMP24]]
1358 //
1359 //
1360 // CHECK9-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1361 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
1362 // CHECK9-NEXT: entry:
1363 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
1364 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1365 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1366 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1367 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1368 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1369 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1370 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1371 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1372 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1373 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1374 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
1375 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1376 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1377 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1378 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1379 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1380 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1381 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1382 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1383 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1384 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1385 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1386 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1387 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
1388 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1389 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1390 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
1391 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1392 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1393 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
1394 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1395 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
1396 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1397 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1398 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1399 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1400 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1401 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1402 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1403 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8
1404 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1405 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1406 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1407 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1408 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1409 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1410 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1411 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1412 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1413 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1414 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1415 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1416 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1417 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1418 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1419 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1420 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1421 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1422 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1423 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1424 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1425 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1426 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1427 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1428 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1429 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4
1430 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1431 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4
1432 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1433 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1434 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1435 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1436 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1437 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1438 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1439 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1440 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1441 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8
1442 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1443 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8
1444 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1445 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1446 // CHECK9-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1447 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1448 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1449 // CHECK9: omp_offload.failed:
1450 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1451 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1452 // CHECK9: omp_offload.cont:
1453 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0
1454 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1455 // CHECK9-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1456 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
1457 // CHECK9-NEXT: ret i32 [[TMP41]]
1458 //
1459 //
1460 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75
1461 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1462 // CHECK9-NEXT: entry:
1463 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1464 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1465 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1466 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1467 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1468 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1469 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1470 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1471 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1472 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1473 // CHECK9-NEXT: ret void
1474 //
1475 //
1476 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1477 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1478 // CHECK9-NEXT: entry:
1479 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1480 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1481 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1482 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1483 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1484 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1485 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1486 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1487 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1488 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1489 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1490 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1491 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1492 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1493 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1494 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1495 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1496 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1497 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1498 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1499 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1500 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1501 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1502 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1503 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1504 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1505 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1506 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1507 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1508 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1509 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1510 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1511 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1512 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1513 // CHECK9: omp.precond.then:
1514 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1515 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1516 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1517 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1518 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1519 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1520 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1521 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1522 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1523 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1524 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1525 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1526 // CHECK9: cond.true:
1527 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1528 // CHECK9-NEXT: br label [[COND_END:%.*]]
1529 // CHECK9: cond.false:
1530 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1531 // CHECK9-NEXT: br label [[COND_END]]
1532 // CHECK9: cond.end:
1533 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1534 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1535 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1536 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1537 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1538 // CHECK9: omp.inner.for.cond:
1539 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
1540 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4
1541 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1542 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1543 // CHECK9: omp.inner.for.body:
1544 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
1545 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1546 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1547 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !4
1548 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !4
1549 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1550 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1551 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !4
1552 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1553 // CHECK9: omp.body.continue:
1554 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1555 // CHECK9: omp.inner.for.inc:
1556 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
1557 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1558 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
1559 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
1560 // CHECK9: omp.inner.for.end:
1561 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1562 // CHECK9: omp.loop.exit:
1563 // CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1564 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1565 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1566 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1567 // CHECK9-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
1568 // CHECK9-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1569 // CHECK9: .omp.final.then:
1570 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1571 // CHECK9-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0
1572 // CHECK9-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
1573 // CHECK9-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
1574 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
1575 // CHECK9-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
1576 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
1577 // CHECK9: .omp.final.done:
1578 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1579 // CHECK9: omp.precond.end:
1580 // CHECK9-NEXT: ret void
1581 //
1582 //
1583 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1584 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
1585 // CHECK9-NEXT: entry:
1586 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
1587 // CHECK9-NEXT: ret void
1588 //
1589 //
1590 // CHECK11-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1591 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
1592 // CHECK11-NEXT: entry:
1593 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
1594 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1595 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1596 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1597 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1598 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1599 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1600 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
1601 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1602 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1603 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1604 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
1605 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1606 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
1607 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
1608 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
1609 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1610 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
1611 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
1612 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
1613 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
1614 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
1615 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1616 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
1617 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1618 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
1619 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
1620 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1621 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
1622 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
1623 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1624 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
1625 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1626 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
1627 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
1628 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1629 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1630 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
1631 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1632 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
1633 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1634 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1635 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
1636 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1637 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1638 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
1639 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1640 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
1641 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1642 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
1643 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1644 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1645 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1646 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1647 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1648 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1649 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1650 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1651 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1652 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1653 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1654 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1655 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1656 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1657 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1658 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4
1659 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1660 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4
1661 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1662 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
1663 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1664 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
1665 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1666 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
1667 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1668 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
1669 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1670 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4
1671 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1672 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4
1673 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1674 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1675 // CHECK11-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1676 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1677 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1678 // CHECK11: omp_offload.failed:
1679 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1680 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
1681 // CHECK11: omp_offload.cont:
1682 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0
1683 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1684 // CHECK11-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
1685 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
1686 // CHECK11-NEXT: ret i32 [[TMP41]]
1687 //
1688 //
1689 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l75
1690 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1691 // CHECK11-NEXT: entry:
1692 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1693 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1694 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1695 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1696 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1697 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1698 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1699 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1700 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
1701 // CHECK11-NEXT: ret void
1702 //
1703 //
1704 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1705 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1706 // CHECK11-NEXT: entry:
1707 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1708 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1709 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1710 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1711 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1712 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1713 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1714 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1715 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1716 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
1717 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1718 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1719 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1720 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1721 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
1722 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1723 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1724 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1725 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1726 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1727 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1728 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1729 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1730 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1731 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1732 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1733 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1734 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1735 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1736 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1737 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
1738 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1739 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1740 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1741 // CHECK11: omp.precond.then:
1742 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1743 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1744 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1745 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1746 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1747 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1748 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1749 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1750 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1751 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1752 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1753 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1754 // CHECK11: cond.true:
1755 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1756 // CHECK11-NEXT: br label [[COND_END:%.*]]
1757 // CHECK11: cond.false:
1758 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1759 // CHECK11-NEXT: br label [[COND_END]]
1760 // CHECK11: cond.end:
1761 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1762 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1763 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1764 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1765 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1766 // CHECK11: omp.inner.for.cond:
1767 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
1768 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
1769 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1770 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1771 // CHECK11: omp.inner.for.body:
1772 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
1773 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1774 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1775 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !5
1776 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5
1777 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
1778 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
1779 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1780 // CHECK11: omp.body.continue:
1781 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1782 // CHECK11: omp.inner.for.inc:
1783 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
1784 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1785 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
1786 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
1787 // CHECK11: omp.inner.for.end:
1788 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1789 // CHECK11: omp.loop.exit:
1790 // CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1791 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1792 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1793 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1794 // CHECK11-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
1795 // CHECK11-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1796 // CHECK11: .omp.final.then:
1797 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1798 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0
1799 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
1800 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
1801 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
1802 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
1803 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
1804 // CHECK11: .omp.final.done:
1805 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
1806 // CHECK11: omp.precond.end:
1807 // CHECK11-NEXT: ret void
1808 //
1809 //
1810 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1811 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] {
1812 // CHECK11-NEXT: entry:
1813 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
1814 // CHECK11-NEXT: ret void
1815 //
1816 //
1817 // CHECK13-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1818 // CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
1819 // CHECK13-NEXT: entry:
1820 // CHECK13-NEXT: [[N:%.*]] = alloca i32, align 4
1821 // CHECK13-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1822 // CHECK13-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1823 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
1824 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1825 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1826 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1827 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1828 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
1829 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1830 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
1831 // CHECK13-NEXT: store i32 100, i32* [[N]], align 4
1832 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1833 // CHECK13-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1834 // CHECK13-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1835 // CHECK13-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1836 // CHECK13-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1837 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1838 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1839 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1840 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1841 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1842 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1843 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1844 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1845 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1846 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1847 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1848 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
1849 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1850 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
1851 // CHECK13-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
1852 // CHECK13: simd.if.then:
1853 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1854 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1855 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1856 // CHECK13: omp.inner.for.cond:
1857 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1858 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1859 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1860 // CHECK13-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1861 // CHECK13: omp.inner.for.body:
1862 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1863 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1864 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1865 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
1866 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
1867 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1868 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
1869 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
1870 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1871 // CHECK13: omp.body.continue:
1872 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1873 // CHECK13: omp.inner.for.inc:
1874 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1875 // CHECK13-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
1876 // CHECK13-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1877 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1878 // CHECK13: omp.inner.for.end:
1879 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1880 // CHECK13-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
1881 // CHECK13-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1882 // CHECK13-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
1883 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
1884 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
1885 // CHECK13-NEXT: br label [[SIMD_IF_END]]
1886 // CHECK13: simd.if.end:
1887 // CHECK13-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0
1888 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[ARRAYIDX10]], align 4
1889 // CHECK13-NEXT: [[TMP15:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1890 // CHECK13-NEXT: call void @llvm.stackrestore(i8* [[TMP15]])
1891 // CHECK13-NEXT: ret i32 [[TMP14]]
1892 //
1893 //
1894 // CHECK15-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1895 // CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
1896 // CHECK15-NEXT: entry:
1897 // CHECK15-NEXT: [[N:%.*]] = alloca i32, align 4
1898 // CHECK15-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1899 // CHECK15-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1900 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
1901 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1902 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1903 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1904 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1905 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
1906 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1907 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
1908 // CHECK15-NEXT: store i32 100, i32* [[N]], align 4
1909 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1910 // CHECK15-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
1911 // CHECK15-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
1912 // CHECK15-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
1913 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1914 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
1915 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1916 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1917 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1918 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1919 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1920 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1921 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1922 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1923 // CHECK15-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
1924 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
1925 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1926 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1927 // CHECK15-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
1928 // CHECK15: simd.if.then:
1929 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1930 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1931 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1932 // CHECK15: omp.inner.for.cond:
1933 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1934 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
1935 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1936 // CHECK15-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1937 // CHECK15: omp.inner.for.body:
1938 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1939 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1940 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1941 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
1942 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
1943 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
1944 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
1945 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1946 // CHECK15: omp.body.continue:
1947 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1948 // CHECK15: omp.inner.for.inc:
1949 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1950 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
1951 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1952 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
1953 // CHECK15: omp.inner.for.end:
1954 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1955 // CHECK15-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
1956 // CHECK15-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1957 // CHECK15-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
1958 // CHECK15-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
1959 // CHECK15-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
1960 // CHECK15-NEXT: br label [[SIMD_IF_END]]
1961 // CHECK15: simd.if.end:
1962 // CHECK15-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0
1963 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[ARRAYIDX10]], align 4
1964 // CHECK15-NEXT: [[TMP14:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
1965 // CHECK15-NEXT: call void @llvm.stackrestore(i8* [[TMP14]])
1966 // CHECK15-NEXT: ret i32 [[TMP13]]
1967 //
1968 //
1969 // CHECK17-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1970 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] {
1971 // CHECK17-NEXT: entry:
1972 // CHECK17-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1973 // CHECK17-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
1974 // CHECK17-NEXT: ret i32 [[CALL]]
1975 //
1976 //
1977 // CHECK17-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1978 // CHECK17-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1979 // CHECK17-NEXT: entry:
1980 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1981 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1982 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1983 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1984 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1985 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
1986 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1987 // CHECK17-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1988 // CHECK17-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
1989 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1990 // CHECK17-NEXT: [[TMP0:%.*]] = getelementptr float, float* [[B]], i32 1
1991 // CHECK17-NEXT: [[TMP1:%.*]] = bitcast [123 x i32]* [[A]] to i8*
1992 // CHECK17-NEXT: [[TMP2:%.*]] = bitcast float* [[TMP0]] to i8*
1993 // CHECK17-NEXT: [[TMP3:%.*]] = ptrtoint i8* [[TMP2]] to i64
1994 // CHECK17-NEXT: [[TMP4:%.*]] = ptrtoint i8* [[TMP1]] to i64
1995 // CHECK17-NEXT: [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
1996 // CHECK17-NEXT: [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1997 // CHECK17-NEXT: [[TMP7:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1998 // CHECK17-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
1999 // CHECK17-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2000 // CHECK17-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to %struct.SS**
2001 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP9]], align 8
2002 // CHECK17-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2003 // CHECK17-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [123 x i32]**
2004 // CHECK17-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP11]], align 8
2005 // CHECK17-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2006 // CHECK17-NEXT: store i64 [[TMP6]], i64* [[TMP12]], align 8
2007 // CHECK17-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2008 // CHECK17-NEXT: store i8* null, i8** [[TMP13]], align 8
2009 // CHECK17-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2010 // CHECK17-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.SS**
2011 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP15]], align 8
2012 // CHECK17-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2013 // CHECK17-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to float**
2014 // CHECK17-NEXT: store float* [[B]], float** [[TMP17]], align 8
2015 // CHECK17-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2016 // CHECK17-NEXT: store i8* null, i8** [[TMP18]], align 8
2017 // CHECK17-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2018 // CHECK17-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to %struct.SS**
2019 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP20]], align 8
2020 // CHECK17-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2021 // CHECK17-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [123 x i32]**
2022 // CHECK17-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP22]], align 8
2023 // CHECK17-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2024 // CHECK17-NEXT: store i8* null, i8** [[TMP23]], align 8
2025 // CHECK17-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2026 // CHECK17-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2027 // CHECK17-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2028 // CHECK17-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2029 // CHECK17-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2030 // CHECK17-NEXT: store i32 1, i32* [[TMP27]], align 4
2031 // CHECK17-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2032 // CHECK17-NEXT: store i32 3, i32* [[TMP28]], align 4
2033 // CHECK17-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2034 // CHECK17-NEXT: store i8** [[TMP24]], i8*** [[TMP29]], align 8
2035 // CHECK17-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2036 // CHECK17-NEXT: store i8** [[TMP25]], i8*** [[TMP30]], align 8
2037 // CHECK17-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2038 // CHECK17-NEXT: store i64* [[TMP26]], i64** [[TMP31]], align 8
2039 // CHECK17-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2040 // CHECK17-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 8
2041 // CHECK17-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2042 // CHECK17-NEXT: store i8** null, i8*** [[TMP33]], align 8
2043 // CHECK17-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2044 // CHECK17-NEXT: store i8** null, i8*** [[TMP34]], align 8
2045 // CHECK17-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2046 // CHECK17-NEXT: store i64 123, i64* [[TMP35]], align 8
2047 // CHECK17-NEXT: [[TMP36:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2048 // CHECK17-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2049 // CHECK17-NEXT: br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2050 // CHECK17: omp_offload.failed:
2051 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2052 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]]
2053 // CHECK17: omp_offload.cont:
2054 // CHECK17-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2055 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0
2056 // CHECK17-NEXT: [[TMP38:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2057 // CHECK17-NEXT: ret i32 [[TMP38]]
2058 //
2059 //
2060 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123
2061 // CHECK17-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2062 // CHECK17-NEXT: entry:
2063 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2064 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2065 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2066 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2067 // CHECK17-NEXT: ret void
2068 //
2069 //
2070 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
2071 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
2072 // CHECK17-NEXT: entry:
2073 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2074 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2075 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2076 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2077 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
2078 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2079 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2080 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2081 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2082 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
2083 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2084 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2085 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2086 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2087 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2088 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2089 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2090 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2091 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2092 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2093 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2094 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2095 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2096 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2097 // CHECK17: cond.true:
2098 // CHECK17-NEXT: br label [[COND_END:%.*]]
2099 // CHECK17: cond.false:
2100 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2101 // CHECK17-NEXT: br label [[COND_END]]
2102 // CHECK17: cond.end:
2103 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2104 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2105 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2106 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2107 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2108 // CHECK17: omp.inner.for.cond:
2109 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2110 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4
2111 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2112 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2113 // CHECK17: omp.inner.for.body:
2114 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2115 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2116 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2117 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !4
2118 // CHECK17-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2119 // CHECK17-NEXT: [[TMP9:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !4
2120 // CHECK17-NEXT: [[CONV:%.*]] = fptosi float [[TMP9]] to i32
2121 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2122 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4
2123 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP10]] to i64
2124 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2125 // CHECK17-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !4
2126 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2127 // CHECK17: omp.body.continue:
2128 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2129 // CHECK17: omp.inner.for.inc:
2130 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2131 // CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
2132 // CHECK17-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2133 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
2134 // CHECK17: omp.inner.for.end:
2135 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2136 // CHECK17: omp.loop.exit:
2137 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2138 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2139 // CHECK17-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2140 // CHECK17-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2141 // CHECK17: .omp.final.then:
2142 // CHECK17-NEXT: store i32 123, i32* [[I]], align 4
2143 // CHECK17-NEXT: br label [[DOTOMP_FINAL_DONE]]
2144 // CHECK17: .omp.final.done:
2145 // CHECK17-NEXT: ret void
2146 //
2147 //
2148 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2149 // CHECK17-SAME: () #[[ATTR4:[0-9]+]] {
2150 // CHECK17-NEXT: entry:
2151 // CHECK17-NEXT: call void @__tgt_register_requires(i64 1)
2152 // CHECK17-NEXT: ret void
2153 //
2154 //
2155 // CHECK19-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2156 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] {
2157 // CHECK19-NEXT: entry:
2158 // CHECK19-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2159 // CHECK19-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2160 // CHECK19-NEXT: ret i32 [[CALL]]
2161 //
2162 //
2163 // CHECK19-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2164 // CHECK19-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2165 // CHECK19-NEXT: entry:
2166 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2167 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2168 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2169 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2170 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2171 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
2172 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2173 // CHECK19-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2174 // CHECK19-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2175 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2176 // CHECK19-NEXT: [[TMP0:%.*]] = getelementptr float, float* [[B]], i32 1
2177 // CHECK19-NEXT: [[TMP1:%.*]] = bitcast [123 x i32]* [[A]] to i8*
2178 // CHECK19-NEXT: [[TMP2:%.*]] = bitcast float* [[TMP0]] to i8*
2179 // CHECK19-NEXT: [[TMP3:%.*]] = ptrtoint i8* [[TMP2]] to i64
2180 // CHECK19-NEXT: [[TMP4:%.*]] = ptrtoint i8* [[TMP1]] to i64
2181 // CHECK19-NEXT: [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
2182 // CHECK19-NEXT: [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
2183 // CHECK19-NEXT: [[TMP7:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2184 // CHECK19-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP7]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
2185 // CHECK19-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2186 // CHECK19-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to %struct.SS**
2187 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP9]], align 4
2188 // CHECK19-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2189 // CHECK19-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [123 x i32]**
2190 // CHECK19-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP11]], align 4
2191 // CHECK19-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2192 // CHECK19-NEXT: store i64 [[TMP6]], i64* [[TMP12]], align 4
2193 // CHECK19-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2194 // CHECK19-NEXT: store i8* null, i8** [[TMP13]], align 4
2195 // CHECK19-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2196 // CHECK19-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.SS**
2197 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP15]], align 4
2198 // CHECK19-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2199 // CHECK19-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to float**
2200 // CHECK19-NEXT: store float* [[B]], float** [[TMP17]], align 4
2201 // CHECK19-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2202 // CHECK19-NEXT: store i8* null, i8** [[TMP18]], align 4
2203 // CHECK19-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2204 // CHECK19-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to %struct.SS**
2205 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP20]], align 4
2206 // CHECK19-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2207 // CHECK19-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [123 x i32]**
2208 // CHECK19-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP22]], align 4
2209 // CHECK19-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2210 // CHECK19-NEXT: store i8* null, i8** [[TMP23]], align 4
2211 // CHECK19-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2212 // CHECK19-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2213 // CHECK19-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2214 // CHECK19-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2215 // CHECK19-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2216 // CHECK19-NEXT: store i32 1, i32* [[TMP27]], align 4
2217 // CHECK19-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2218 // CHECK19-NEXT: store i32 3, i32* [[TMP28]], align 4
2219 // CHECK19-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2220 // CHECK19-NEXT: store i8** [[TMP24]], i8*** [[TMP29]], align 4
2221 // CHECK19-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2222 // CHECK19-NEXT: store i8** [[TMP25]], i8*** [[TMP30]], align 4
2223 // CHECK19-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2224 // CHECK19-NEXT: store i64* [[TMP26]], i64** [[TMP31]], align 4
2225 // CHECK19-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2226 // CHECK19-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 4
2227 // CHECK19-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2228 // CHECK19-NEXT: store i8** null, i8*** [[TMP33]], align 4
2229 // CHECK19-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2230 // CHECK19-NEXT: store i8** null, i8*** [[TMP34]], align 4
2231 // CHECK19-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2232 // CHECK19-NEXT: store i64 123, i64* [[TMP35]], align 8
2233 // CHECK19-NEXT: [[TMP36:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2234 // CHECK19-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2235 // CHECK19-NEXT: br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2236 // CHECK19: omp_offload.failed:
2237 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2238 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]]
2239 // CHECK19: omp_offload.cont:
2240 // CHECK19-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2241 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0
2242 // CHECK19-NEXT: [[TMP38:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2243 // CHECK19-NEXT: ret i32 [[TMP38]]
2244 //
2245 //
2246 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123
2247 // CHECK19-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2248 // CHECK19-NEXT: entry:
2249 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2250 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2251 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2252 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2253 // CHECK19-NEXT: ret void
2254 //
2255 //
2256 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
2257 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
2258 // CHECK19-NEXT: entry:
2259 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2260 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2261 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2262 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2263 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
2264 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2265 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2266 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2267 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2268 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
2269 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2270 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2271 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2272 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2273 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2274 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2275 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2276 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2277 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2278 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2279 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2280 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2281 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2282 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2283 // CHECK19: cond.true:
2284 // CHECK19-NEXT: br label [[COND_END:%.*]]
2285 // CHECK19: cond.false:
2286 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2287 // CHECK19-NEXT: br label [[COND_END]]
2288 // CHECK19: cond.end:
2289 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2290 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2291 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2292 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2293 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2294 // CHECK19: omp.inner.for.cond:
2295 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2296 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
2297 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2298 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2299 // CHECK19: omp.inner.for.body:
2300 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2301 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2302 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2303 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !5
2304 // CHECK19-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2305 // CHECK19-NEXT: [[TMP9:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !5
2306 // CHECK19-NEXT: [[CONV:%.*]] = fptosi float [[TMP9]] to i32
2307 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2308 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5
2309 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP10]]
2310 // CHECK19-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
2311 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2312 // CHECK19: omp.body.continue:
2313 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2314 // CHECK19: omp.inner.for.inc:
2315 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2316 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], 1
2317 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2318 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
2319 // CHECK19: omp.inner.for.end:
2320 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2321 // CHECK19: omp.loop.exit:
2322 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2323 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2324 // CHECK19-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2325 // CHECK19-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2326 // CHECK19: .omp.final.then:
2327 // CHECK19-NEXT: store i32 123, i32* [[I]], align 4
2328 // CHECK19-NEXT: br label [[DOTOMP_FINAL_DONE]]
2329 // CHECK19: .omp.final.done:
2330 // CHECK19-NEXT: ret void
2331 //
2332 //
2333 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2334 // CHECK19-SAME: () #[[ATTR4:[0-9]+]] {
2335 // CHECK19-NEXT: entry:
2336 // CHECK19-NEXT: call void @__tgt_register_requires(i64 1)
2337 // CHECK19-NEXT: ret void
2338 //
2339 //
2340 // CHECK21-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2341 // CHECK21-SAME: () #[[ATTR0:[0-9]+]] {
2342 // CHECK21-NEXT: entry:
2343 // CHECK21-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2344 // CHECK21-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2345 // CHECK21-NEXT: ret i32 [[CALL]]
2346 //
2347 //
2348 // CHECK21-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2349 // CHECK21-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2350 // CHECK21-NEXT: entry:
2351 // CHECK21-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2352 // CHECK21-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2353 // CHECK21-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2354 // CHECK21-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2355 // CHECK21-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
2356 // CHECK21-NEXT: [[TMP:%.*]] = alloca i32, align 4
2357 // CHECK21-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2358 // CHECK21-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2359 // CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2360 // CHECK21-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2361 // CHECK21-NEXT: [[TMP0:%.*]] = getelementptr float, float* [[B]], i32 1
2362 // CHECK21-NEXT: [[TMP1:%.*]] = bitcast [123 x i32]* [[A]] to i8*
2363 // CHECK21-NEXT: [[TMP2:%.*]] = bitcast float* [[TMP0]] to i8*
2364 // CHECK21-NEXT: [[TMP3:%.*]] = ptrtoint i8* [[TMP2]] to i64
2365 // CHECK21-NEXT: [[TMP4:%.*]] = ptrtoint i8* [[TMP1]] to i64
2366 // CHECK21-NEXT: [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
2367 // CHECK21-NEXT: [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
2368 // CHECK21-NEXT: [[TMP7:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2369 // CHECK21-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP7]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
2370 // CHECK21-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2371 // CHECK21-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to %struct.SS**
2372 // CHECK21-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP9]], align 8
2373 // CHECK21-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2374 // CHECK21-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [123 x i32]**
2375 // CHECK21-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP11]], align 8
2376 // CHECK21-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2377 // CHECK21-NEXT: store i64 [[TMP6]], i64* [[TMP12]], align 8
2378 // CHECK21-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2379 // CHECK21-NEXT: store i8* null, i8** [[TMP13]], align 8
2380 // CHECK21-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2381 // CHECK21-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.SS**
2382 // CHECK21-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP15]], align 8
2383 // CHECK21-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2384 // CHECK21-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to float**
2385 // CHECK21-NEXT: store float* [[B]], float** [[TMP17]], align 8
2386 // CHECK21-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2387 // CHECK21-NEXT: store i8* null, i8** [[TMP18]], align 8
2388 // CHECK21-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2389 // CHECK21-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to %struct.SS**
2390 // CHECK21-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP20]], align 8
2391 // CHECK21-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2392 // CHECK21-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [123 x i32]**
2393 // CHECK21-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP22]], align 8
2394 // CHECK21-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2395 // CHECK21-NEXT: store i8* null, i8** [[TMP23]], align 8
2396 // CHECK21-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2397 // CHECK21-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2398 // CHECK21-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2399 // CHECK21-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2400 // CHECK21-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2401 // CHECK21-NEXT: store i32 1, i32* [[TMP27]], align 4
2402 // CHECK21-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2403 // CHECK21-NEXT: store i32 3, i32* [[TMP28]], align 4
2404 // CHECK21-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2405 // CHECK21-NEXT: store i8** [[TMP24]], i8*** [[TMP29]], align 8
2406 // CHECK21-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2407 // CHECK21-NEXT: store i8** [[TMP25]], i8*** [[TMP30]], align 8
2408 // CHECK21-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2409 // CHECK21-NEXT: store i64* [[TMP26]], i64** [[TMP31]], align 8
2410 // CHECK21-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2411 // CHECK21-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 8
2412 // CHECK21-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2413 // CHECK21-NEXT: store i8** null, i8*** [[TMP33]], align 8
2414 // CHECK21-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2415 // CHECK21-NEXT: store i8** null, i8*** [[TMP34]], align 8
2416 // CHECK21-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2417 // CHECK21-NEXT: store i64 123, i64* [[TMP35]], align 8
2418 // CHECK21-NEXT: [[TMP36:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2419 // CHECK21-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2420 // CHECK21-NEXT: br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2421 // CHECK21: omp_offload.failed:
2422 // CHECK21-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2423 // CHECK21-NEXT: br label [[OMP_OFFLOAD_CONT]]
2424 // CHECK21: omp_offload.cont:
2425 // CHECK21-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2426 // CHECK21-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0
2427 // CHECK21-NEXT: [[TMP38:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2428 // CHECK21-NEXT: ret i32 [[TMP38]]
2429 //
2430 //
2431 // CHECK21-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123
2432 // CHECK21-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2433 // CHECK21-NEXT: entry:
2434 // CHECK21-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2435 // CHECK21-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2436 // CHECK21-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2437 // CHECK21-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2438 // CHECK21-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2439 // CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2440 // CHECK21-NEXT: [[TMP1:%.*]] = load float, float* [[B]], align 4
2441 // CHECK21-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
2442 // CHECK21-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
2443 // CHECK21-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
2444 // CHECK21-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2445 // CHECK21-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP2]] to i1
2446 // CHECK21-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
2447 // CHECK21-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8
2448 // CHECK21-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1
2449 // CHECK21-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2450 // CHECK21-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]], i64 [[TMP3]])
2451 // CHECK21-NEXT: ret void
2452 //
2453 //
2454 // CHECK21-LABEL: define {{[^@]+}}@.omp_outlined.
2455 // CHECK21-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
2456 // CHECK21-NEXT: entry:
2457 // CHECK21-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2458 // CHECK21-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2459 // CHECK21-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2460 // CHECK21-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2461 // CHECK21-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2462 // CHECK21-NEXT: [[TMP:%.*]] = alloca i32, align 4
2463 // CHECK21-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2464 // CHECK21-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2465 // CHECK21-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2466 // CHECK21-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2467 // CHECK21-NEXT: [[I:%.*]] = alloca i32, align 4
2468 // CHECK21-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2469 // CHECK21-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2470 // CHECK21-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2471 // CHECK21-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2472 // CHECK21-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2473 // CHECK21-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
2474 // CHECK21-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2475 // CHECK21-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2476 // CHECK21-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2477 // CHECK21-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2478 // CHECK21-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2479 // CHECK21-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2480 // CHECK21-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2481 // CHECK21-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2482 // CHECK21-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2483 // CHECK21-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2484 // CHECK21: cond.true:
2485 // CHECK21-NEXT: br label [[COND_END:%.*]]
2486 // CHECK21: cond.false:
2487 // CHECK21-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2488 // CHECK21-NEXT: br label [[COND_END]]
2489 // CHECK21: cond.end:
2490 // CHECK21-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2491 // CHECK21-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2492 // CHECK21-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2493 // CHECK21-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2494 // CHECK21-NEXT: [[TMP6:%.*]] = load i8, i8* [[CONV]], align 1
2495 // CHECK21-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP6]] to i1
2496 // CHECK21-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2497 // CHECK21: omp_if.then:
2498 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2499 // CHECK21: omp.inner.for.cond:
2500 // CHECK21-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2501 // CHECK21-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4
2502 // CHECK21-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2503 // CHECK21-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2504 // CHECK21: omp.inner.for.body:
2505 // CHECK21-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2506 // CHECK21-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2507 // CHECK21-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2508 // CHECK21-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !4
2509 // CHECK21-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2510 // CHECK21-NEXT: [[TMP10:%.*]] = load float, float* [[B]], align 4, !nontemporal !5, !llvm.access.group !4
2511 // CHECK21-NEXT: [[CONV2:%.*]] = fptosi float [[TMP10]] to i32
2512 // CHECK21-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2513 // CHECK21-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4
2514 // CHECK21-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2515 // CHECK21-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2516 // CHECK21-NEXT: store i32 [[CONV2]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !4
2517 // CHECK21-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2518 // CHECK21: omp.body.continue:
2519 // CHECK21-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2520 // CHECK21: omp.inner.for.inc:
2521 // CHECK21-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2522 // CHECK21-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2523 // CHECK21-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
2524 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
2525 // CHECK21: omp.inner.for.end:
2526 // CHECK21-NEXT: br label [[OMP_IF_END:%.*]]
2527 // CHECK21: omp_if.else:
2528 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]]
2529 // CHECK21: omp.inner.for.cond4:
2530 // CHECK21-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2531 // CHECK21-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2532 // CHECK21-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2533 // CHECK21-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY6:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
2534 // CHECK21: omp.inner.for.body6:
2535 // CHECK21-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2536 // CHECK21-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP15]], 1
2537 // CHECK21-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL7]]
2538 // CHECK21-NEXT: store i32 [[ADD8]], i32* [[I]], align 4
2539 // CHECK21-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 1
2540 // CHECK21-NEXT: [[TMP16:%.*]] = load float, float* [[B9]], align 4
2541 // CHECK21-NEXT: [[CONV10:%.*]] = fptosi float [[TMP16]] to i32
2542 // CHECK21-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2543 // CHECK21-NEXT: [[TMP17:%.*]] = load i32, i32* [[I]], align 4
2544 // CHECK21-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP17]] to i64
2545 // CHECK21-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A11]], i64 0, i64 [[IDXPROM12]]
2546 // CHECK21-NEXT: store i32 [[CONV10]], i32* [[ARRAYIDX13]], align 4
2547 // CHECK21-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
2548 // CHECK21: omp.body.continue14:
2549 // CHECK21-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
2550 // CHECK21: omp.inner.for.inc15:
2551 // CHECK21-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2552 // CHECK21-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP18]], 1
2553 // CHECK21-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4
2554 // CHECK21-NEXT: br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP9:![0-9]+]]
2555 // CHECK21: omp.inner.for.end17:
2556 // CHECK21-NEXT: br label [[OMP_IF_END]]
2557 // CHECK21: omp_if.end:
2558 // CHECK21-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2559 // CHECK21: omp.loop.exit:
2560 // CHECK21-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2561 // CHECK21-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2562 // CHECK21-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
2563 // CHECK21-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2564 // CHECK21: .omp.final.then:
2565 // CHECK21-NEXT: store i32 123, i32* [[I]], align 4
2566 // CHECK21-NEXT: br label [[DOTOMP_FINAL_DONE]]
2567 // CHECK21: .omp.final.done:
2568 // CHECK21-NEXT: ret void
2569 //
2570 //
2571 // CHECK21-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2572 // CHECK21-SAME: () #[[ATTR4:[0-9]+]] {
2573 // CHECK21-NEXT: entry:
2574 // CHECK21-NEXT: call void @__tgt_register_requires(i64 1)
2575 // CHECK21-NEXT: ret void
2576 //
2577 //
2578 // CHECK23-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2579 // CHECK23-SAME: () #[[ATTR0:[0-9]+]] {
2580 // CHECK23-NEXT: entry:
2581 // CHECK23-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2582 // CHECK23-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2583 // CHECK23-NEXT: ret i32 [[CALL]]
2584 //
2585 //
2586 // CHECK23-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2587 // CHECK23-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2588 // CHECK23-NEXT: entry:
2589 // CHECK23-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2590 // CHECK23-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2591 // CHECK23-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2592 // CHECK23-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2593 // CHECK23-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2594 // CHECK23-NEXT: [[TMP:%.*]] = alloca i32, align 4
2595 // CHECK23-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2596 // CHECK23-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2597 // CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2598 // CHECK23-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2599 // CHECK23-NEXT: [[TMP0:%.*]] = getelementptr float, float* [[B]], i32 1
2600 // CHECK23-NEXT: [[TMP1:%.*]] = bitcast [123 x i32]* [[A]] to i8*
2601 // CHECK23-NEXT: [[TMP2:%.*]] = bitcast float* [[TMP0]] to i8*
2602 // CHECK23-NEXT: [[TMP3:%.*]] = ptrtoint i8* [[TMP2]] to i64
2603 // CHECK23-NEXT: [[TMP4:%.*]] = ptrtoint i8* [[TMP1]] to i64
2604 // CHECK23-NEXT: [[TMP5:%.*]] = sub i64 [[TMP3]], [[TMP4]]
2605 // CHECK23-NEXT: [[TMP6:%.*]] = sdiv exact i64 [[TMP5]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
2606 // CHECK23-NEXT: [[TMP7:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2607 // CHECK23-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP7]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
2608 // CHECK23-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2609 // CHECK23-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to %struct.SS**
2610 // CHECK23-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP9]], align 4
2611 // CHECK23-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2612 // CHECK23-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to [123 x i32]**
2613 // CHECK23-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP11]], align 4
2614 // CHECK23-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2615 // CHECK23-NEXT: store i64 [[TMP6]], i64* [[TMP12]], align 4
2616 // CHECK23-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2617 // CHECK23-NEXT: store i8* null, i8** [[TMP13]], align 4
2618 // CHECK23-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2619 // CHECK23-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to %struct.SS**
2620 // CHECK23-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP15]], align 4
2621 // CHECK23-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2622 // CHECK23-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to float**
2623 // CHECK23-NEXT: store float* [[B]], float** [[TMP17]], align 4
2624 // CHECK23-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2625 // CHECK23-NEXT: store i8* null, i8** [[TMP18]], align 4
2626 // CHECK23-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2627 // CHECK23-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to %struct.SS**
2628 // CHECK23-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP20]], align 4
2629 // CHECK23-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2630 // CHECK23-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to [123 x i32]**
2631 // CHECK23-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP22]], align 4
2632 // CHECK23-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2633 // CHECK23-NEXT: store i8* null, i8** [[TMP23]], align 4
2634 // CHECK23-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2635 // CHECK23-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2636 // CHECK23-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2637 // CHECK23-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2638 // CHECK23-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2639 // CHECK23-NEXT: store i32 1, i32* [[TMP27]], align 4
2640 // CHECK23-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2641 // CHECK23-NEXT: store i32 3, i32* [[TMP28]], align 4
2642 // CHECK23-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2643 // CHECK23-NEXT: store i8** [[TMP24]], i8*** [[TMP29]], align 4
2644 // CHECK23-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2645 // CHECK23-NEXT: store i8** [[TMP25]], i8*** [[TMP30]], align 4
2646 // CHECK23-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2647 // CHECK23-NEXT: store i64* [[TMP26]], i64** [[TMP31]], align 4
2648 // CHECK23-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2649 // CHECK23-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP32]], align 4
2650 // CHECK23-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2651 // CHECK23-NEXT: store i8** null, i8*** [[TMP33]], align 4
2652 // CHECK23-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2653 // CHECK23-NEXT: store i8** null, i8*** [[TMP34]], align 4
2654 // CHECK23-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2655 // CHECK23-NEXT: store i64 123, i64* [[TMP35]], align 8
2656 // CHECK23-NEXT: [[TMP36:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2657 // CHECK23-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2658 // CHECK23-NEXT: br i1 [[TMP37]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2659 // CHECK23: omp_offload.failed:
2660 // CHECK23-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2661 // CHECK23-NEXT: br label [[OMP_OFFLOAD_CONT]]
2662 // CHECK23: omp_offload.cont:
2663 // CHECK23-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2664 // CHECK23-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0
2665 // CHECK23-NEXT: [[TMP38:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2666 // CHECK23-NEXT: ret i32 [[TMP38]]
2667 //
2668 //
2669 // CHECK23-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l123
2670 // CHECK23-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2671 // CHECK23-NEXT: entry:
2672 // CHECK23-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2673 // CHECK23-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2674 // CHECK23-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2675 // CHECK23-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2676 // CHECK23-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2677 // CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2678 // CHECK23-NEXT: [[TMP1:%.*]] = load float, float* [[B]], align 4
2679 // CHECK23-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP1]], 0.000000e+00
2680 // CHECK23-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
2681 // CHECK23-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
2682 // CHECK23-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2683 // CHECK23-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP2]] to i1
2684 // CHECK23-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
2685 // CHECK23-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8
2686 // CHECK23-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1
2687 // CHECK23-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2688 // CHECK23-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]], i32 [[TMP3]])
2689 // CHECK23-NEXT: ret void
2690 //
2691 //
2692 // CHECK23-LABEL: define {{[^@]+}}@.omp_outlined.
2693 // CHECK23-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
2694 // CHECK23-NEXT: entry:
2695 // CHECK23-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2696 // CHECK23-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2697 // CHECK23-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2698 // CHECK23-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2699 // CHECK23-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2700 // CHECK23-NEXT: [[TMP:%.*]] = alloca i32, align 4
2701 // CHECK23-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2702 // CHECK23-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2703 // CHECK23-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2704 // CHECK23-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2705 // CHECK23-NEXT: [[I:%.*]] = alloca i32, align 4
2706 // CHECK23-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2707 // CHECK23-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2708 // CHECK23-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2709 // CHECK23-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2710 // CHECK23-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2711 // CHECK23-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
2712 // CHECK23-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2713 // CHECK23-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2714 // CHECK23-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2715 // CHECK23-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2716 // CHECK23-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2717 // CHECK23-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2718 // CHECK23-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2719 // CHECK23-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2720 // CHECK23-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2721 // CHECK23-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2722 // CHECK23: cond.true:
2723 // CHECK23-NEXT: br label [[COND_END:%.*]]
2724 // CHECK23: cond.false:
2725 // CHECK23-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2726 // CHECK23-NEXT: br label [[COND_END]]
2727 // CHECK23: cond.end:
2728 // CHECK23-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2729 // CHECK23-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2730 // CHECK23-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2731 // CHECK23-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2732 // CHECK23-NEXT: [[TMP6:%.*]] = load i8, i8* [[CONV]], align 1
2733 // CHECK23-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP6]] to i1
2734 // CHECK23-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2735 // CHECK23: omp_if.then:
2736 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2737 // CHECK23: omp.inner.for.cond:
2738 // CHECK23-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2739 // CHECK23-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
2740 // CHECK23-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2741 // CHECK23-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2742 // CHECK23: omp.inner.for.body:
2743 // CHECK23-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2744 // CHECK23-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2745 // CHECK23-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2746 // CHECK23-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !5
2747 // CHECK23-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 1
2748 // CHECK23-NEXT: [[TMP10:%.*]] = load float, float* [[B]], align 4, !nontemporal !6, !llvm.access.group !5
2749 // CHECK23-NEXT: [[CONV2:%.*]] = fptosi float [[TMP10]] to i32
2750 // CHECK23-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2751 // CHECK23-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5
2752 // CHECK23-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2753 // CHECK23-NEXT: store i32 [[CONV2]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
2754 // CHECK23-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2755 // CHECK23: omp.body.continue:
2756 // CHECK23-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2757 // CHECK23: omp.inner.for.inc:
2758 // CHECK23-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2759 // CHECK23-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2760 // CHECK23-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
2761 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
2762 // CHECK23: omp.inner.for.end:
2763 // CHECK23-NEXT: br label [[OMP_IF_END:%.*]]
2764 // CHECK23: omp_if.else:
2765 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND4:%.*]]
2766 // CHECK23: omp.inner.for.cond4:
2767 // CHECK23-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2768 // CHECK23-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2769 // CHECK23-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2770 // CHECK23-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY6:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
2771 // CHECK23: omp.inner.for.body6:
2772 // CHECK23-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2773 // CHECK23-NEXT: [[MUL7:%.*]] = mul nsw i32 [[TMP15]], 1
2774 // CHECK23-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL7]]
2775 // CHECK23-NEXT: store i32 [[ADD8]], i32* [[I]], align 4
2776 // CHECK23-NEXT: [[B9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 1
2777 // CHECK23-NEXT: [[TMP16:%.*]] = load float, float* [[B9]], align 4
2778 // CHECK23-NEXT: [[CONV10:%.*]] = fptosi float [[TMP16]] to i32
2779 // CHECK23-NEXT: [[A11:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[TMP0]], i32 0, i32 0
2780 // CHECK23-NEXT: [[TMP17:%.*]] = load i32, i32* [[I]], align 4
2781 // CHECK23-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A11]], i32 0, i32 [[TMP17]]
2782 // CHECK23-NEXT: store i32 [[CONV10]], i32* [[ARRAYIDX12]], align 4
2783 // CHECK23-NEXT: br label [[OMP_BODY_CONTINUE13:%.*]]
2784 // CHECK23: omp.body.continue13:
2785 // CHECK23-NEXT: br label [[OMP_INNER_FOR_INC14:%.*]]
2786 // CHECK23: omp.inner.for.inc14:
2787 // CHECK23-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2788 // CHECK23-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP18]], 1
2789 // CHECK23-NEXT: store i32 [[ADD15]], i32* [[DOTOMP_IV]], align 4
2790 // CHECK23-NEXT: br label [[OMP_INNER_FOR_COND4]], !llvm.loop [[LOOP10:![0-9]+]]
2791 // CHECK23: omp.inner.for.end16:
2792 // CHECK23-NEXT: br label [[OMP_IF_END]]
2793 // CHECK23: omp_if.end:
2794 // CHECK23-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2795 // CHECK23: omp.loop.exit:
2796 // CHECK23-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2797 // CHECK23-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2798 // CHECK23-NEXT: [[TMP20:%.*]] = icmp ne i32 [[TMP19]], 0
2799 // CHECK23-NEXT: br i1 [[TMP20]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2800 // CHECK23: .omp.final.then:
2801 // CHECK23-NEXT: store i32 123, i32* [[I]], align 4
2802 // CHECK23-NEXT: br label [[DOTOMP_FINAL_DONE]]
2803 // CHECK23: .omp.final.done:
2804 // CHECK23-NEXT: ret void
2805 //
2806 //
2807 // CHECK23-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2808 // CHECK23-SAME: () #[[ATTR4:[0-9]+]] {
2809 // CHECK23-NEXT: entry:
2810 // CHECK23-NEXT: call void @__tgt_register_requires(i64 1)
2811 // CHECK23-NEXT: ret void
2812 //
2813 //
2814 // CHECK25-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2815 // CHECK25-SAME: () #[[ATTR0:[0-9]+]] {
2816 // CHECK25-NEXT: entry:
2817 // CHECK25-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2818 // CHECK25-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2819 // CHECK25-NEXT: ret i32 [[CALL]]
2820 //
2821 //
2822 // CHECK25-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2823 // CHECK25-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2824 // CHECK25-NEXT: entry:
2825 // CHECK25-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2826 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
2827 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2828 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2829 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2830 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4
2831 // CHECK25-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2832 // CHECK25-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2833 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2834 // CHECK25-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2835 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2836 // CHECK25-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2837 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2838 // CHECK25: omp.inner.for.cond:
2839 // CHECK25-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2840 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
2841 // CHECK25-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2842 // CHECK25-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2843 // CHECK25: omp.inner.for.body:
2844 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2845 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2846 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2847 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
2848 // CHECK25-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2849 // CHECK25-NEXT: [[TMP4:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !2
2850 // CHECK25-NEXT: [[CONV:%.*]] = fptosi float [[TMP4]] to i32
2851 // CHECK25-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2852 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
2853 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP5]] to i64
2854 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2855 // CHECK25-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
2856 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2857 // CHECK25: omp.body.continue:
2858 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2859 // CHECK25: omp.inner.for.inc:
2860 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2861 // CHECK25-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP6]], 1
2862 // CHECK25-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2863 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
2864 // CHECK25: omp.inner.for.end:
2865 // CHECK25-NEXT: store i32 123, i32* [[I]], align 4
2866 // CHECK25-NEXT: [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2867 // CHECK25-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A3]], i64 0, i64 0
2868 // CHECK25-NEXT: [[TMP7:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
2869 // CHECK25-NEXT: ret i32 [[TMP7]]
2870 //
2871 //
2872 // CHECK27-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2873 // CHECK27-SAME: () #[[ATTR0:[0-9]+]] {
2874 // CHECK27-NEXT: entry:
2875 // CHECK27-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2876 // CHECK27-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2877 // CHECK27-NEXT: ret i32 [[CALL]]
2878 //
2879 //
2880 // CHECK27-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2881 // CHECK27-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2882 // CHECK27-NEXT: entry:
2883 // CHECK27-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2884 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
2885 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2886 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2887 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2888 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4
2889 // CHECK27-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2890 // CHECK27-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2891 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2892 // CHECK27-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2893 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2894 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2895 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2896 // CHECK27: omp.inner.for.cond:
2897 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2898 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
2899 // CHECK27-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2900 // CHECK27-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2901 // CHECK27: omp.inner.for.body:
2902 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2903 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2904 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2905 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
2906 // CHECK27-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2907 // CHECK27-NEXT: [[TMP4:%.*]] = load float, float* [[B]], align 4, !llvm.access.group !3
2908 // CHECK27-NEXT: [[CONV:%.*]] = fptosi float [[TMP4]] to i32
2909 // CHECK27-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2910 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
2911 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP5]]
2912 // CHECK27-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
2913 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2914 // CHECK27: omp.body.continue:
2915 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2916 // CHECK27: omp.inner.for.inc:
2917 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2918 // CHECK27-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP6]], 1
2919 // CHECK27-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2920 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
2921 // CHECK27: omp.inner.for.end:
2922 // CHECK27-NEXT: store i32 123, i32* [[I]], align 4
2923 // CHECK27-NEXT: [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2924 // CHECK27-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A3]], i32 0, i32 0
2925 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
2926 // CHECK27-NEXT: ret i32 [[TMP7]]
2927 //
2928 //
2929 // CHECK29-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2930 // CHECK29-SAME: () #[[ATTR0:[0-9]+]] {
2931 // CHECK29-NEXT: entry:
2932 // CHECK29-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2933 // CHECK29-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
2934 // CHECK29-NEXT: ret i32 [[CALL]]
2935 //
2936 //
2937 // CHECK29-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2938 // CHECK29-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2939 // CHECK29-NEXT: entry:
2940 // CHECK29-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2941 // CHECK29-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2942 // CHECK29-NEXT: [[TMP:%.*]] = alloca i32, align 4
2943 // CHECK29-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2944 // CHECK29-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2945 // CHECK29-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2946 // CHECK29-NEXT: [[I:%.*]] = alloca i32, align 4
2947 // CHECK29-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2948 // CHECK29-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2949 // CHECK29-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
2950 // CHECK29-NEXT: [[TMP0:%.*]] = load float, float* [[B]], align 4
2951 // CHECK29-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP0]], 0.000000e+00
2952 // CHECK29-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
2953 // CHECK29-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
2954 // CHECK29-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2955 // CHECK29-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2956 // CHECK29-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2957 // CHECK29-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
2958 // CHECK29-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2959 // CHECK29-NEXT: [[TOBOOL2:%.*]] = trunc i8 [[TMP2]] to i1
2960 // CHECK29-NEXT: br i1 [[TOBOOL2]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2961 // CHECK29: omp_if.then:
2962 // CHECK29-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2963 // CHECK29: omp.inner.for.cond:
2964 // CHECK29-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2965 // CHECK29-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
2966 // CHECK29-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
2967 // CHECK29-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2968 // CHECK29: omp.inner.for.body:
2969 // CHECK29-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2970 // CHECK29-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP5]], 1
2971 // CHECK29-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2972 // CHECK29-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
2973 // CHECK29-NEXT: [[B3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 1
2974 // CHECK29-NEXT: [[TMP6:%.*]] = load float, float* [[B3]], align 4, !nontemporal !3, !llvm.access.group !2
2975 // CHECK29-NEXT: [[CONV:%.*]] = fptosi float [[TMP6]] to i32
2976 // CHECK29-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2977 // CHECK29-NEXT: [[TMP7:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
2978 // CHECK29-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP7]] to i64
2979 // CHECK29-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2980 // CHECK29-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
2981 // CHECK29-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2982 // CHECK29: omp.body.continue:
2983 // CHECK29-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2984 // CHECK29: omp.inner.for.inc:
2985 // CHECK29-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2986 // CHECK29-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
2987 // CHECK29-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2988 // CHECK29-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
2989 // CHECK29: omp.inner.for.end:
2990 // CHECK29-NEXT: br label [[OMP_IF_END:%.*]]
2991 // CHECK29: omp_if.else:
2992 // CHECK29-NEXT: br label [[OMP_INNER_FOR_COND5:%.*]]
2993 // CHECK29: omp.inner.for.cond5:
2994 // CHECK29-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2995 // CHECK29-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2996 // CHECK29-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
2997 // CHECK29-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2998 // CHECK29: omp.inner.for.body7:
2999 // CHECK29-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3000 // CHECK29-NEXT: [[MUL8:%.*]] = mul nsw i32 [[TMP11]], 1
3001 // CHECK29-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
3002 // CHECK29-NEXT: store i32 [[ADD9]], i32* [[I]], align 4
3003 // CHECK29-NEXT: [[B10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 1
3004 // CHECK29-NEXT: [[TMP12:%.*]] = load float, float* [[B10]], align 4
3005 // CHECK29-NEXT: [[CONV11:%.*]] = fptosi float [[TMP12]] to i32
3006 // CHECK29-NEXT: [[A12:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3007 // CHECK29-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4
3008 // CHECK29-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP13]] to i64
3009 // CHECK29-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A12]], i64 0, i64 [[IDXPROM13]]
3010 // CHECK29-NEXT: store i32 [[CONV11]], i32* [[ARRAYIDX14]], align 4
3011 // CHECK29-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
3012 // CHECK29: omp.body.continue15:
3013 // CHECK29-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
3014 // CHECK29: omp.inner.for.inc16:
3015 // CHECK29-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3016 // CHECK29-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP14]], 1
3017 // CHECK29-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
3018 // CHECK29-NEXT: br label [[OMP_INNER_FOR_COND5]], !llvm.loop [[LOOP7:![0-9]+]]
3019 // CHECK29: omp.inner.for.end18:
3020 // CHECK29-NEXT: br label [[OMP_IF_END]]
3021 // CHECK29: omp_if.end:
3022 // CHECK29-NEXT: store i32 123, i32* [[I]], align 4
3023 // CHECK29-NEXT: [[A19:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3024 // CHECK29-NEXT: [[ARRAYIDX20:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A19]], i64 0, i64 0
3025 // CHECK29-NEXT: [[TMP15:%.*]] = load i32, i32* [[ARRAYIDX20]], align 4
3026 // CHECK29-NEXT: ret i32 [[TMP15]]
3027 //
3028 //
3029 // CHECK31-LABEL: define {{[^@]+}}@_Z21teams_template_structv
3030 // CHECK31-SAME: () #[[ATTR0:[0-9]+]] {
3031 // CHECK31-NEXT: entry:
3032 // CHECK31-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
3033 // CHECK31-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
3034 // CHECK31-NEXT: ret i32 [[CALL]]
3035 //
3036 //
3037 // CHECK31-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
3038 // CHECK31-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
3039 // CHECK31-NEXT: entry:
3040 // CHECK31-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3041 // CHECK31-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3042 // CHECK31-NEXT: [[TMP:%.*]] = alloca i32, align 4
3043 // CHECK31-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3044 // CHECK31-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3045 // CHECK31-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3046 // CHECK31-NEXT: [[I:%.*]] = alloca i32, align 4
3047 // CHECK31-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3048 // CHECK31-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3049 // CHECK31-NEXT: [[B:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 1
3050 // CHECK31-NEXT: [[TMP0:%.*]] = load float, float* [[B]], align 4
3051 // CHECK31-NEXT: [[TOBOOL:%.*]] = fcmp une float [[TMP0]], 0.000000e+00
3052 // CHECK31-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
3053 // CHECK31-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
3054 // CHECK31-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3055 // CHECK31-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3056 // CHECK31-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3057 // CHECK31-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_IV]], align 4
3058 // CHECK31-NEXT: [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3059 // CHECK31-NEXT: [[TOBOOL2:%.*]] = trunc i8 [[TMP2]] to i1
3060 // CHECK31-NEXT: br i1 [[TOBOOL2]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3061 // CHECK31: omp_if.then:
3062 // CHECK31-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3063 // CHECK31: omp.inner.for.cond:
3064 // CHECK31-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
3065 // CHECK31-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
3066 // CHECK31-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP3]], [[TMP4]]
3067 // CHECK31-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3068 // CHECK31: omp.inner.for.body:
3069 // CHECK31-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
3070 // CHECK31-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP5]], 1
3071 // CHECK31-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3072 // CHECK31-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
3073 // CHECK31-NEXT: [[B3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 1
3074 // CHECK31-NEXT: [[TMP6:%.*]] = load float, float* [[B3]], align 4, !nontemporal !4, !llvm.access.group !3
3075 // CHECK31-NEXT: [[CONV:%.*]] = fptosi float [[TMP6]] to i32
3076 // CHECK31-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3077 // CHECK31-NEXT: [[TMP7:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
3078 // CHECK31-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP7]]
3079 // CHECK31-NEXT: store i32 [[CONV]], i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
3080 // CHECK31-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3081 // CHECK31: omp.body.continue:
3082 // CHECK31-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3083 // CHECK31: omp.inner.for.inc:
3084 // CHECK31-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
3085 // CHECK31-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP8]], 1
3086 // CHECK31-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
3087 // CHECK31-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
3088 // CHECK31: omp.inner.for.end:
3089 // CHECK31-NEXT: br label [[OMP_IF_END:%.*]]
3090 // CHECK31: omp_if.else:
3091 // CHECK31-NEXT: br label [[OMP_INNER_FOR_COND5:%.*]]
3092 // CHECK31: omp.inner.for.cond5:
3093 // CHECK31-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3094 // CHECK31-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3095 // CHECK31-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
3096 // CHECK31-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY7:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
3097 // CHECK31: omp.inner.for.body7:
3098 // CHECK31-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3099 // CHECK31-NEXT: [[MUL8:%.*]] = mul nsw i32 [[TMP11]], 1
3100 // CHECK31-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
3101 // CHECK31-NEXT: store i32 [[ADD9]], i32* [[I]], align 4
3102 // CHECK31-NEXT: [[B10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 1
3103 // CHECK31-NEXT: [[TMP12:%.*]] = load float, float* [[B10]], align 4
3104 // CHECK31-NEXT: [[CONV11:%.*]] = fptosi float [[TMP12]] to i32
3105 // CHECK31-NEXT: [[A12:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3106 // CHECK31-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4
3107 // CHECK31-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A12]], i32 0, i32 [[TMP13]]
3108 // CHECK31-NEXT: store i32 [[CONV11]], i32* [[ARRAYIDX13]], align 4
3109 // CHECK31-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
3110 // CHECK31: omp.body.continue14:
3111 // CHECK31-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
3112 // CHECK31: omp.inner.for.inc15:
3113 // CHECK31-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3114 // CHECK31-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP14]], 1
3115 // CHECK31-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4
3116 // CHECK31-NEXT: br label [[OMP_INNER_FOR_COND5]], !llvm.loop [[LOOP8:![0-9]+]]
3117 // CHECK31: omp.inner.for.end17:
3118 // CHECK31-NEXT: br label [[OMP_IF_END]]
3119 // CHECK31: omp_if.end:
3120 // CHECK31-NEXT: store i32 123, i32* [[I]], align 4
3121 // CHECK31-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3122 // CHECK31-NEXT: [[ARRAYIDX19:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i32 0, i32 0
3123 // CHECK31-NEXT: [[TMP15:%.*]] = load i32, i32* [[ARRAYIDX19]], align 4
3124 // CHECK31-NEXT: ret i32 [[TMP15]]
3125 //
3126 //
3127 // CHECK33-LABEL: define {{[^@]+}}@main
3128 // CHECK33-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3129 // CHECK33-NEXT: entry:
3130 // CHECK33-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3131 // CHECK33-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3132 // CHECK33-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3133 // CHECK33-NEXT: [[N:%.*]] = alloca i32, align 4
3134 // CHECK33-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
3135 // CHECK33-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3136 // CHECK33-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
3137 // CHECK33-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3138 // CHECK33-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3139 // CHECK33-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3140 // CHECK33-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
3141 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4
3142 // CHECK33-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3143 // CHECK33-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3144 // CHECK33-NEXT: store i32 0, i32* [[RETVAL]], align 4
3145 // CHECK33-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3146 // CHECK33-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3147 // CHECK33-NEXT: store i32 100, i32* [[N]], align 4
3148 // CHECK33-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3149 // CHECK33-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
3150 // CHECK33-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
3151 // CHECK33-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
3152 // CHECK33-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
3153 // CHECK33-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
3154 // CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
3155 // CHECK33-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3156 // CHECK33-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
3157 // CHECK33-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
3158 // CHECK33-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
3159 // CHECK33-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3160 // CHECK33-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
3161 // CHECK33-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3162 // CHECK33-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
3163 // CHECK33-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
3164 // CHECK33-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3165 // CHECK33-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3166 // CHECK33-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
3167 // CHECK33-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3168 // CHECK33-NEXT: store i8* null, i8** [[TMP11]], align 8
3169 // CHECK33-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3170 // CHECK33-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
3171 // CHECK33-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
3172 // CHECK33-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3173 // CHECK33-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3174 // CHECK33-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
3175 // CHECK33-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3176 // CHECK33-NEXT: store i8* null, i8** [[TMP16]], align 8
3177 // CHECK33-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3178 // CHECK33-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
3179 // CHECK33-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
3180 // CHECK33-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3181 // CHECK33-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
3182 // CHECK33-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
3183 // CHECK33-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3184 // CHECK33-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
3185 // CHECK33-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3186 // CHECK33-NEXT: store i8* null, i8** [[TMP22]], align 8
3187 // CHECK33-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3188 // CHECK33-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3189 // CHECK33-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3190 // CHECK33-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
3191 // CHECK33-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
3192 // CHECK33-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3193 // CHECK33-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
3194 // CHECK33-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3195 // CHECK33-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3196 // CHECK33-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3197 // CHECK33-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3198 // CHECK33-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
3199 // CHECK33-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
3200 // CHECK33-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3201 // CHECK33-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3202 // CHECK33-NEXT: store i32 1, i32* [[TMP30]], align 4
3203 // CHECK33-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3204 // CHECK33-NEXT: store i32 3, i32* [[TMP31]], align 4
3205 // CHECK33-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3206 // CHECK33-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
3207 // CHECK33-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3208 // CHECK33-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
3209 // CHECK33-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3210 // CHECK33-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
3211 // CHECK33-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3212 // CHECK33-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
3213 // CHECK33-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3214 // CHECK33-NEXT: store i8** null, i8*** [[TMP36]], align 8
3215 // CHECK33-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3216 // CHECK33-NEXT: store i8** null, i8*** [[TMP37]], align 8
3217 // CHECK33-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3218 // CHECK33-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
3219 // CHECK33-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3220 // CHECK33-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
3221 // CHECK33-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3222 // CHECK33: omp_offload.failed:
3223 // CHECK33-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3224 // CHECK33-NEXT: br label [[OMP_OFFLOAD_CONT]]
3225 // CHECK33: omp_offload.cont:
3226 // CHECK33-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3227 // CHECK33-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP41]])
3228 // CHECK33-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3229 // CHECK33-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3230 // CHECK33-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
3231 // CHECK33-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4
3232 // CHECK33-NEXT: ret i32 [[TMP43]]
3233 //
3234 //
3235 // CHECK33-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192
3236 // CHECK33-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3237 // CHECK33-NEXT: entry:
3238 // CHECK33-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3239 // CHECK33-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3240 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3241 // CHECK33-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3242 // CHECK33-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3243 // CHECK33-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3244 // CHECK33-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3245 // CHECK33-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3246 // CHECK33-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3247 // CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
3248 // CHECK33-NEXT: ret void
3249 //
3250 //
3251 // CHECK33-LABEL: define {{[^@]+}}@.omp_outlined.
3252 // CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3253 // CHECK33-NEXT: entry:
3254 // CHECK33-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3255 // CHECK33-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3256 // CHECK33-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3257 // CHECK33-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3258 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3259 // CHECK33-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3260 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4
3261 // CHECK33-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3262 // CHECK33-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3263 // CHECK33-NEXT: [[I:%.*]] = alloca i32, align 4
3264 // CHECK33-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3265 // CHECK33-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3266 // CHECK33-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3267 // CHECK33-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3268 // CHECK33-NEXT: [[I3:%.*]] = alloca i32, align 4
3269 // CHECK33-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3270 // CHECK33-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3271 // CHECK33-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3272 // CHECK33-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3273 // CHECK33-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3274 // CHECK33-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3275 // CHECK33-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3276 // CHECK33-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3277 // CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3278 // CHECK33-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3279 // CHECK33-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3280 // CHECK33-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3281 // CHECK33-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3282 // CHECK33-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3283 // CHECK33-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3284 // CHECK33-NEXT: store i32 0, i32* [[I]], align 4
3285 // CHECK33-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3286 // CHECK33-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3287 // CHECK33-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3288 // CHECK33: omp.precond.then:
3289 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3290 // CHECK33-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3291 // CHECK33-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3292 // CHECK33-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3293 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3294 // CHECK33-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3295 // CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3296 // CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3297 // CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3298 // CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3299 // CHECK33-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3300 // CHECK33-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3301 // CHECK33: cond.true:
3302 // CHECK33-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3303 // CHECK33-NEXT: br label [[COND_END:%.*]]
3304 // CHECK33: cond.false:
3305 // CHECK33-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3306 // CHECK33-NEXT: br label [[COND_END]]
3307 // CHECK33: cond.end:
3308 // CHECK33-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3309 // CHECK33-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3310 // CHECK33-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3311 // CHECK33-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3312 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3313 // CHECK33: omp.inner.for.cond:
3314 // CHECK33-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
3315 // CHECK33-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
3316 // CHECK33-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3317 // CHECK33-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3318 // CHECK33: omp.inner.for.body:
3319 // CHECK33-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
3320 // CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
3321 // CHECK33-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3322 // CHECK33-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !5
3323 // CHECK33-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !5
3324 // CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
3325 // CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
3326 // CHECK33-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
3327 // CHECK33-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3328 // CHECK33: omp.body.continue:
3329 // CHECK33-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3330 // CHECK33: omp.inner.for.inc:
3331 // CHECK33-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
3332 // CHECK33-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
3333 // CHECK33-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
3334 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
3335 // CHECK33: omp.inner.for.end:
3336 // CHECK33-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3337 // CHECK33: omp.loop.exit:
3338 // CHECK33-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3339 // CHECK33-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3340 // CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
3341 // CHECK33-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3342 // CHECK33-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
3343 // CHECK33-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3344 // CHECK33: .omp.final.then:
3345 // CHECK33-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3346 // CHECK33-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0
3347 // CHECK33-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
3348 // CHECK33-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
3349 // CHECK33-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
3350 // CHECK33-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
3351 // CHECK33-NEXT: br label [[DOTOMP_FINAL_DONE]]
3352 // CHECK33: .omp.final.done:
3353 // CHECK33-NEXT: br label [[OMP_PRECOND_END]]
3354 // CHECK33: omp.precond.end:
3355 // CHECK33-NEXT: ret void
3356 //
3357 //
3358 // CHECK33-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3359 // CHECK33-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
3360 // CHECK33-NEXT: entry:
3361 // CHECK33-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3362 // CHECK33-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3363 // CHECK33-NEXT: [[TE:%.*]] = alloca i32, align 4
3364 // CHECK33-NEXT: [[TH:%.*]] = alloca i32, align 4
3365 // CHECK33-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
3366 // CHECK33-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
3367 // CHECK33-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3368 // CHECK33-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3369 // CHECK33-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3370 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4
3371 // CHECK33-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3372 // CHECK33-NEXT: store i32 0, i32* [[TE]], align 4
3373 // CHECK33-NEXT: store i32 128, i32* [[TH]], align 4
3374 // CHECK33-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
3375 // CHECK33-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
3376 // CHECK33-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
3377 // CHECK33-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8
3378 // CHECK33-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
3379 // CHECK33-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
3380 // CHECK33-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
3381 // CHECK33-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8
3382 // CHECK33-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3383 // CHECK33-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
3384 // CHECK33-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8
3385 // CHECK33-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3386 // CHECK33-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3387 // CHECK33-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8
3388 // CHECK33-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3389 // CHECK33-NEXT: store i8* null, i8** [[TMP8]], align 8
3390 // CHECK33-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3391 // CHECK33-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3392 // CHECK33-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8
3393 // CHECK33-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3394 // CHECK33-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3395 // CHECK33-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
3396 // CHECK33-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3397 // CHECK33-NEXT: store i8* null, i8** [[TMP13]], align 8
3398 // CHECK33-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3399 // CHECK33-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
3400 // CHECK33-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8
3401 // CHECK33-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3402 // CHECK33-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
3403 // CHECK33-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8
3404 // CHECK33-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3405 // CHECK33-NEXT: store i8* null, i8** [[TMP18]], align 8
3406 // CHECK33-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3407 // CHECK33-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3408 // CHECK33-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
3409 // CHECK33-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3410 // CHECK33-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3411 // CHECK33-NEXT: store i32 1, i32* [[TMP22]], align 4
3412 // CHECK33-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3413 // CHECK33-NEXT: store i32 3, i32* [[TMP23]], align 4
3414 // CHECK33-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3415 // CHECK33-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 8
3416 // CHECK33-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3417 // CHECK33-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 8
3418 // CHECK33-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3419 // CHECK33-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP26]], align 8
3420 // CHECK33-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3421 // CHECK33-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP27]], align 8
3422 // CHECK33-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3423 // CHECK33-NEXT: store i8** null, i8*** [[TMP28]], align 8
3424 // CHECK33-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3425 // CHECK33-NEXT: store i8** null, i8*** [[TMP29]], align 8
3426 // CHECK33-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3427 // CHECK33-NEXT: store i64 10, i64* [[TMP30]], align 8
3428 // CHECK33-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3429 // CHECK33-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3430 // CHECK33-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3431 // CHECK33: omp_offload.failed:
3432 // CHECK33-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
3433 // CHECK33-NEXT: br label [[OMP_OFFLOAD_CONT]]
3434 // CHECK33: omp_offload.cont:
3435 // CHECK33-NEXT: ret i32 0
3436 //
3437 //
3438 // CHECK33-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181
3439 // CHECK33-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3440 // CHECK33-NEXT: entry:
3441 // CHECK33-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
3442 // CHECK33-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
3443 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3444 // CHECK33-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
3445 // CHECK33-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
3446 // CHECK33-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
3447 // CHECK33-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3448 // CHECK33-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
3449 // CHECK33-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
3450 // CHECK33-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3451 // CHECK33-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
3452 // CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
3453 // CHECK33-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3454 // CHECK33-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
3455 // CHECK33-NEXT: ret void
3456 //
3457 //
3458 // CHECK33-LABEL: define {{[^@]+}}@.omp_outlined..1
3459 // CHECK33-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3460 // CHECK33-NEXT: entry:
3461 // CHECK33-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3462 // CHECK33-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3463 // CHECK33-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3464 // CHECK33-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3465 // CHECK33-NEXT: [[TMP:%.*]] = alloca i32, align 4
3466 // CHECK33-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3467 // CHECK33-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3468 // CHECK33-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3469 // CHECK33-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3470 // CHECK33-NEXT: [[I:%.*]] = alloca i32, align 4
3471 // CHECK33-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3472 // CHECK33-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3473 // CHECK33-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3474 // CHECK33-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3475 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3476 // CHECK33-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3477 // CHECK33-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3478 // CHECK33-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3479 // CHECK33-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3480 // CHECK33-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3481 // CHECK33-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3482 // CHECK33-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3483 // CHECK33-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3484 // CHECK33-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3485 // CHECK33: cond.true:
3486 // CHECK33-NEXT: br label [[COND_END:%.*]]
3487 // CHECK33: cond.false:
3488 // CHECK33-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3489 // CHECK33-NEXT: br label [[COND_END]]
3490 // CHECK33: cond.end:
3491 // CHECK33-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3492 // CHECK33-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3493 // CHECK33-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3494 // CHECK33-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3495 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3496 // CHECK33: omp.inner.for.cond:
3497 // CHECK33-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3498 // CHECK33-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
3499 // CHECK33-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3500 // CHECK33-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3501 // CHECK33: omp.inner.for.body:
3502 // CHECK33-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3503 // CHECK33-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3504 // CHECK33-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3505 // CHECK33-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
3506 // CHECK33-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
3507 // CHECK33-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
3508 // CHECK33-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3509 // CHECK33-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
3510 // CHECK33-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3511 // CHECK33: omp.body.continue:
3512 // CHECK33-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3513 // CHECK33: omp.inner.for.inc:
3514 // CHECK33-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3515 // CHECK33-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
3516 // CHECK33-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3517 // CHECK33-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
3518 // CHECK33: omp.inner.for.end:
3519 // CHECK33-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3520 // CHECK33: omp.loop.exit:
3521 // CHECK33-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3522 // CHECK33-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3523 // CHECK33-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3524 // CHECK33-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3525 // CHECK33: .omp.final.then:
3526 // CHECK33-NEXT: store i32 10, i32* [[I]], align 4
3527 // CHECK33-NEXT: br label [[DOTOMP_FINAL_DONE]]
3528 // CHECK33: .omp.final.done:
3529 // CHECK33-NEXT: ret void
3530 //
3531 //
3532 // CHECK33-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3533 // CHECK33-SAME: () #[[ATTR6:[0-9]+]] {
3534 // CHECK33-NEXT: entry:
3535 // CHECK33-NEXT: call void @__tgt_register_requires(i64 1)
3536 // CHECK33-NEXT: ret void
3537 //
3538 //
3539 // CHECK35-LABEL: define {{[^@]+}}@main
3540 // CHECK35-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3541 // CHECK35-NEXT: entry:
3542 // CHECK35-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3543 // CHECK35-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3544 // CHECK35-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
3545 // CHECK35-NEXT: [[N:%.*]] = alloca i32, align 4
3546 // CHECK35-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
3547 // CHECK35-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3548 // CHECK35-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
3549 // CHECK35-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3550 // CHECK35-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3551 // CHECK35-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3552 // CHECK35-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
3553 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4
3554 // CHECK35-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3555 // CHECK35-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3556 // CHECK35-NEXT: store i32 0, i32* [[RETVAL]], align 4
3557 // CHECK35-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3558 // CHECK35-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
3559 // CHECK35-NEXT: store i32 100, i32* [[N]], align 4
3560 // CHECK35-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3561 // CHECK35-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
3562 // CHECK35-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
3563 // CHECK35-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
3564 // CHECK35-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
3565 // CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
3566 // CHECK35-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
3567 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
3568 // CHECK35-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
3569 // CHECK35-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
3570 // CHECK35-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3571 // CHECK35-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
3572 // CHECK35-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3573 // CHECK35-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3574 // CHECK35-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
3575 // CHECK35-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3576 // CHECK35-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3577 // CHECK35-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
3578 // CHECK35-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3579 // CHECK35-NEXT: store i8* null, i8** [[TMP11]], align 4
3580 // CHECK35-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3581 // CHECK35-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3582 // CHECK35-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
3583 // CHECK35-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3584 // CHECK35-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
3585 // CHECK35-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
3586 // CHECK35-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3587 // CHECK35-NEXT: store i8* null, i8** [[TMP16]], align 4
3588 // CHECK35-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3589 // CHECK35-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
3590 // CHECK35-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
3591 // CHECK35-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3592 // CHECK35-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
3593 // CHECK35-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
3594 // CHECK35-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3595 // CHECK35-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
3596 // CHECK35-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3597 // CHECK35-NEXT: store i8* null, i8** [[TMP22]], align 4
3598 // CHECK35-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3599 // CHECK35-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3600 // CHECK35-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3601 // CHECK35-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
3602 // CHECK35-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
3603 // CHECK35-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3604 // CHECK35-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
3605 // CHECK35-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3606 // CHECK35-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3607 // CHECK35-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3608 // CHECK35-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3609 // CHECK35-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
3610 // CHECK35-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
3611 // CHECK35-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3612 // CHECK35-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3613 // CHECK35-NEXT: store i32 1, i32* [[TMP30]], align 4
3614 // CHECK35-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3615 // CHECK35-NEXT: store i32 3, i32* [[TMP31]], align 4
3616 // CHECK35-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3617 // CHECK35-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
3618 // CHECK35-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3619 // CHECK35-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
3620 // CHECK35-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3621 // CHECK35-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
3622 // CHECK35-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3623 // CHECK35-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
3624 // CHECK35-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3625 // CHECK35-NEXT: store i8** null, i8*** [[TMP36]], align 4
3626 // CHECK35-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3627 // CHECK35-NEXT: store i8** null, i8*** [[TMP37]], align 4
3628 // CHECK35-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3629 // CHECK35-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
3630 // CHECK35-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3631 // CHECK35-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
3632 // CHECK35-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3633 // CHECK35: omp_offload.failed:
3634 // CHECK35-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3635 // CHECK35-NEXT: br label [[OMP_OFFLOAD_CONT]]
3636 // CHECK35: omp_offload.cont:
3637 // CHECK35-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3638 // CHECK35-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP41]])
3639 // CHECK35-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3640 // CHECK35-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3641 // CHECK35-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
3642 // CHECK35-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4
3643 // CHECK35-NEXT: ret i32 [[TMP43]]
3644 //
3645 //
3646 // CHECK35-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192
3647 // CHECK35-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3648 // CHECK35-NEXT: entry:
3649 // CHECK35-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3650 // CHECK35-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3651 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3652 // CHECK35-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3653 // CHECK35-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3654 // CHECK35-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3655 // CHECK35-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3656 // CHECK35-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3657 // CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
3658 // CHECK35-NEXT: ret void
3659 //
3660 //
3661 // CHECK35-LABEL: define {{[^@]+}}@.omp_outlined.
3662 // CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3663 // CHECK35-NEXT: entry:
3664 // CHECK35-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3665 // CHECK35-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3666 // CHECK35-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3667 // CHECK35-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3668 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3669 // CHECK35-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3670 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4
3671 // CHECK35-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3672 // CHECK35-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3673 // CHECK35-NEXT: [[I:%.*]] = alloca i32, align 4
3674 // CHECK35-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3675 // CHECK35-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3676 // CHECK35-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3677 // CHECK35-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3678 // CHECK35-NEXT: [[I3:%.*]] = alloca i32, align 4
3679 // CHECK35-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3680 // CHECK35-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3681 // CHECK35-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3682 // CHECK35-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3683 // CHECK35-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3684 // CHECK35-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3685 // CHECK35-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3686 // CHECK35-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3687 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3688 // CHECK35-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3689 // CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3690 // CHECK35-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3691 // CHECK35-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3692 // CHECK35-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3693 // CHECK35-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3694 // CHECK35-NEXT: store i32 0, i32* [[I]], align 4
3695 // CHECK35-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3696 // CHECK35-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3697 // CHECK35-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3698 // CHECK35: omp.precond.then:
3699 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3700 // CHECK35-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3701 // CHECK35-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3702 // CHECK35-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3703 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3704 // CHECK35-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3705 // CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3706 // CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3707 // CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3708 // CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3709 // CHECK35-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3710 // CHECK35-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3711 // CHECK35: cond.true:
3712 // CHECK35-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3713 // CHECK35-NEXT: br label [[COND_END:%.*]]
3714 // CHECK35: cond.false:
3715 // CHECK35-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3716 // CHECK35-NEXT: br label [[COND_END]]
3717 // CHECK35: cond.end:
3718 // CHECK35-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3719 // CHECK35-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3720 // CHECK35-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3721 // CHECK35-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3722 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3723 // CHECK35: omp.inner.for.cond:
3724 // CHECK35-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
3725 // CHECK35-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
3726 // CHECK35-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3727 // CHECK35-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3728 // CHECK35: omp.inner.for.body:
3729 // CHECK35-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
3730 // CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
3731 // CHECK35-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3732 // CHECK35-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !6
3733 // CHECK35-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !6
3734 // CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
3735 // CHECK35-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6
3736 // CHECK35-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3737 // CHECK35: omp.body.continue:
3738 // CHECK35-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3739 // CHECK35: omp.inner.for.inc:
3740 // CHECK35-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
3741 // CHECK35-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
3742 // CHECK35-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
3743 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
3744 // CHECK35: omp.inner.for.end:
3745 // CHECK35-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3746 // CHECK35: omp.loop.exit:
3747 // CHECK35-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3748 // CHECK35-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3749 // CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
3750 // CHECK35-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3751 // CHECK35-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
3752 // CHECK35-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3753 // CHECK35: .omp.final.then:
3754 // CHECK35-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3755 // CHECK35-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP23]], 0
3756 // CHECK35-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
3757 // CHECK35-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
3758 // CHECK35-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
3759 // CHECK35-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
3760 // CHECK35-NEXT: br label [[DOTOMP_FINAL_DONE]]
3761 // CHECK35: .omp.final.done:
3762 // CHECK35-NEXT: br label [[OMP_PRECOND_END]]
3763 // CHECK35: omp.precond.end:
3764 // CHECK35-NEXT: ret void
3765 //
3766 //
3767 // CHECK35-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3768 // CHECK35-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
3769 // CHECK35-NEXT: entry:
3770 // CHECK35-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3771 // CHECK35-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3772 // CHECK35-NEXT: [[TE:%.*]] = alloca i32, align 4
3773 // CHECK35-NEXT: [[TH:%.*]] = alloca i32, align 4
3774 // CHECK35-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
3775 // CHECK35-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
3776 // CHECK35-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3777 // CHECK35-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3778 // CHECK35-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3779 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4
3780 // CHECK35-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3781 // CHECK35-NEXT: store i32 0, i32* [[TE]], align 4
3782 // CHECK35-NEXT: store i32 128, i32* [[TH]], align 4
3783 // CHECK35-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
3784 // CHECK35-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4
3785 // CHECK35-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4
3786 // CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
3787 // CHECK35-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4
3788 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4
3789 // CHECK35-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3790 // CHECK35-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
3791 // CHECK35-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4
3792 // CHECK35-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3793 // CHECK35-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
3794 // CHECK35-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4
3795 // CHECK35-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3796 // CHECK35-NEXT: store i8* null, i8** [[TMP8]], align 4
3797 // CHECK35-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3798 // CHECK35-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3799 // CHECK35-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
3800 // CHECK35-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3801 // CHECK35-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
3802 // CHECK35-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
3803 // CHECK35-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3804 // CHECK35-NEXT: store i8* null, i8** [[TMP13]], align 4
3805 // CHECK35-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3806 // CHECK35-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
3807 // CHECK35-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4
3808 // CHECK35-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3809 // CHECK35-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
3810 // CHECK35-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4
3811 // CHECK35-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3812 // CHECK35-NEXT: store i8* null, i8** [[TMP18]], align 4
3813 // CHECK35-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3814 // CHECK35-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3815 // CHECK35-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
3816 // CHECK35-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3817 // CHECK35-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3818 // CHECK35-NEXT: store i32 1, i32* [[TMP22]], align 4
3819 // CHECK35-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3820 // CHECK35-NEXT: store i32 3, i32* [[TMP23]], align 4
3821 // CHECK35-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3822 // CHECK35-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 4
3823 // CHECK35-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3824 // CHECK35-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 4
3825 // CHECK35-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3826 // CHECK35-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP26]], align 4
3827 // CHECK35-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3828 // CHECK35-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP27]], align 4
3829 // CHECK35-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3830 // CHECK35-NEXT: store i8** null, i8*** [[TMP28]], align 4
3831 // CHECK35-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3832 // CHECK35-NEXT: store i8** null, i8*** [[TMP29]], align 4
3833 // CHECK35-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3834 // CHECK35-NEXT: store i64 10, i64* [[TMP30]], align 8
3835 // CHECK35-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3836 // CHECK35-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3837 // CHECK35-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3838 // CHECK35: omp_offload.failed:
3839 // CHECK35-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
3840 // CHECK35-NEXT: br label [[OMP_OFFLOAD_CONT]]
3841 // CHECK35: omp_offload.cont:
3842 // CHECK35-NEXT: ret i32 0
3843 //
3844 //
3845 // CHECK35-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181
3846 // CHECK35-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3847 // CHECK35-NEXT: entry:
3848 // CHECK35-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
3849 // CHECK35-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
3850 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
3851 // CHECK35-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
3852 // CHECK35-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
3853 // CHECK35-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
3854 // CHECK35-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
3855 // CHECK35-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
3856 // CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
3857 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
3858 // CHECK35-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3859 // CHECK35-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
3860 // CHECK35-NEXT: ret void
3861 //
3862 //
3863 // CHECK35-LABEL: define {{[^@]+}}@.omp_outlined..1
3864 // CHECK35-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3865 // CHECK35-NEXT: entry:
3866 // CHECK35-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3867 // CHECK35-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3868 // CHECK35-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
3869 // CHECK35-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3870 // CHECK35-NEXT: [[TMP:%.*]] = alloca i32, align 4
3871 // CHECK35-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3872 // CHECK35-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3873 // CHECK35-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3874 // CHECK35-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3875 // CHECK35-NEXT: [[I:%.*]] = alloca i32, align 4
3876 // CHECK35-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3877 // CHECK35-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3878 // CHECK35-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
3879 // CHECK35-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
3880 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3881 // CHECK35-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3882 // CHECK35-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3883 // CHECK35-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3884 // CHECK35-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3885 // CHECK35-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3886 // CHECK35-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3887 // CHECK35-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3888 // CHECK35-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3889 // CHECK35-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3890 // CHECK35: cond.true:
3891 // CHECK35-NEXT: br label [[COND_END:%.*]]
3892 // CHECK35: cond.false:
3893 // CHECK35-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3894 // CHECK35-NEXT: br label [[COND_END]]
3895 // CHECK35: cond.end:
3896 // CHECK35-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3897 // CHECK35-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3898 // CHECK35-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3899 // CHECK35-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3900 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3901 // CHECK35: omp.inner.for.cond:
3902 // CHECK35-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
3903 // CHECK35-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
3904 // CHECK35-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3905 // CHECK35-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3906 // CHECK35: omp.inner.for.body:
3907 // CHECK35-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
3908 // CHECK35-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3909 // CHECK35-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3910 // CHECK35-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
3911 // CHECK35-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
3912 // CHECK35-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
3913 // CHECK35-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
3914 // CHECK35-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3915 // CHECK35: omp.body.continue:
3916 // CHECK35-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3917 // CHECK35: omp.inner.for.inc:
3918 // CHECK35-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
3919 // CHECK35-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
3920 // CHECK35-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
3921 // CHECK35-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
3922 // CHECK35: omp.inner.for.end:
3923 // CHECK35-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3924 // CHECK35: omp.loop.exit:
3925 // CHECK35-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3926 // CHECK35-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3927 // CHECK35-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3928 // CHECK35-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3929 // CHECK35: .omp.final.then:
3930 // CHECK35-NEXT: store i32 10, i32* [[I]], align 4
3931 // CHECK35-NEXT: br label [[DOTOMP_FINAL_DONE]]
3932 // CHECK35: .omp.final.done:
3933 // CHECK35-NEXT: ret void
3934 //
3935 //
3936 // CHECK35-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3937 // CHECK35-SAME: () #[[ATTR6:[0-9]+]] {
3938 // CHECK35-NEXT: entry:
3939 // CHECK35-NEXT: call void @__tgt_register_requires(i64 1)
3940 // CHECK35-NEXT: ret void
3941 //
3942 //
3943 // CHECK37-LABEL: define {{[^@]+}}@main
3944 // CHECK37-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3945 // CHECK37-NEXT: entry:
3946 // CHECK37-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3947 // CHECK37-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3948 // CHECK37-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3949 // CHECK37-NEXT: [[N:%.*]] = alloca i32, align 4
3950 // CHECK37-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
3951 // CHECK37-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3952 // CHECK37-NEXT: [[ARGC_CASTED:%.*]] = alloca i64, align 8
3953 // CHECK37-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
3954 // CHECK37-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
3955 // CHECK37-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
3956 // CHECK37-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
3957 // CHECK37-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [4 x i64], align 8
3958 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4
3959 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3960 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3961 // CHECK37-NEXT: store i32 0, i32* [[RETVAL]], align 4
3962 // CHECK37-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3963 // CHECK37-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3964 // CHECK37-NEXT: store i32 100, i32* [[N]], align 4
3965 // CHECK37-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3966 // CHECK37-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
3967 // CHECK37-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
3968 // CHECK37-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
3969 // CHECK37-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
3970 // CHECK37-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
3971 // CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3972 // CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
3973 // CHECK37-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
3974 // CHECK37-NEXT: [[TMP4:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
3975 // CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[N]], align 4
3976 // CHECK37-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3977 // CHECK37-NEXT: store i32 [[TMP5]], i32* [[CONV1]], align 4
3978 // CHECK37-NEXT: [[TMP6:%.*]] = load i64, i64* [[N_CASTED]], align 8
3979 // CHECK37-NEXT: [[TMP7:%.*]] = mul nuw i64 [[TMP1]], 4
3980 // CHECK37-NEXT: [[TMP8:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3981 // CHECK37-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP8]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes to i8*), i64 32, i1 false)
3982 // CHECK37-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3983 // CHECK37-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3984 // CHECK37-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
3985 // CHECK37-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3986 // CHECK37-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3987 // CHECK37-NEXT: store i64 [[TMP4]], i64* [[TMP12]], align 8
3988 // CHECK37-NEXT: [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3989 // CHECK37-NEXT: store i8* null, i8** [[TMP13]], align 8
3990 // CHECK37-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3991 // CHECK37-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3992 // CHECK37-NEXT: store i64 [[TMP6]], i64* [[TMP15]], align 8
3993 // CHECK37-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3994 // CHECK37-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i64*
3995 // CHECK37-NEXT: store i64 [[TMP6]], i64* [[TMP17]], align 8
3996 // CHECK37-NEXT: [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3997 // CHECK37-NEXT: store i8* null, i8** [[TMP18]], align 8
3998 // CHECK37-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3999 // CHECK37-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
4000 // CHECK37-NEXT: store i64 [[TMP1]], i64* [[TMP20]], align 8
4001 // CHECK37-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4002 // CHECK37-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i64*
4003 // CHECK37-NEXT: store i64 [[TMP1]], i64* [[TMP22]], align 8
4004 // CHECK37-NEXT: [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4005 // CHECK37-NEXT: store i8* null, i8** [[TMP23]], align 8
4006 // CHECK37-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4007 // CHECK37-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32**
4008 // CHECK37-NEXT: store i32* [[VLA]], i32** [[TMP25]], align 8
4009 // CHECK37-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4010 // CHECK37-NEXT: [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32**
4011 // CHECK37-NEXT: store i32* [[VLA]], i32** [[TMP27]], align 8
4012 // CHECK37-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
4013 // CHECK37-NEXT: store i64 [[TMP7]], i64* [[TMP28]], align 8
4014 // CHECK37-NEXT: [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
4015 // CHECK37-NEXT: store i8* null, i8** [[TMP29]], align 8
4016 // CHECK37-NEXT: [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4017 // CHECK37-NEXT: [[TMP31:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4018 // CHECK37-NEXT: [[TMP32:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4019 // CHECK37-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
4020 // CHECK37-NEXT: store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_]], align 4
4021 // CHECK37-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4022 // CHECK37-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
4023 // CHECK37-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4024 // CHECK37-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
4025 // CHECK37-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4026 // CHECK37-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4027 // CHECK37-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
4028 // CHECK37-NEXT: [[TMP36:%.*]] = zext i32 [[ADD]] to i64
4029 // CHECK37-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4030 // CHECK37-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
4031 // CHECK37-NEXT: store i32 1, i32* [[TMP37]], align 4
4032 // CHECK37-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
4033 // CHECK37-NEXT: store i32 4, i32* [[TMP38]], align 4
4034 // CHECK37-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
4035 // CHECK37-NEXT: store i8** [[TMP30]], i8*** [[TMP39]], align 8
4036 // CHECK37-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
4037 // CHECK37-NEXT: store i8** [[TMP31]], i8*** [[TMP40]], align 8
4038 // CHECK37-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
4039 // CHECK37-NEXT: store i64* [[TMP32]], i64** [[TMP41]], align 8
4040 // CHECK37-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
4041 // CHECK37-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP42]], align 8
4042 // CHECK37-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
4043 // CHECK37-NEXT: store i8** null, i8*** [[TMP43]], align 8
4044 // CHECK37-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
4045 // CHECK37-NEXT: store i8** null, i8*** [[TMP44]], align 8
4046 // CHECK37-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
4047 // CHECK37-NEXT: store i64 [[TMP36]], i64* [[TMP45]], align 8
4048 // CHECK37-NEXT: [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
4049 // CHECK37-NEXT: [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
4050 // CHECK37-NEXT: br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4051 // CHECK37: omp_offload.failed:
4052 // CHECK37-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192(i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
4053 // CHECK37-NEXT: br label [[OMP_OFFLOAD_CONT]]
4054 // CHECK37: omp_offload.cont:
4055 // CHECK37-NEXT: [[TMP48:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4056 // CHECK37-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP48]])
4057 // CHECK37-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
4058 // CHECK37-NEXT: [[TMP49:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4059 // CHECK37-NEXT: call void @llvm.stackrestore(i8* [[TMP49]])
4060 // CHECK37-NEXT: [[TMP50:%.*]] = load i32, i32* [[RETVAL]], align 4
4061 // CHECK37-NEXT: ret i32 [[TMP50]]
4062 //
4063 //
4064 // CHECK37-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192
4065 // CHECK37-SAME: (i64 noundef [[ARGC:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
4066 // CHECK37-NEXT: entry:
4067 // CHECK37-NEXT: [[ARGC_ADDR:%.*]] = alloca i64, align 8
4068 // CHECK37-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
4069 // CHECK37-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
4070 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
4071 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4072 // CHECK37-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4073 // CHECK37-NEXT: store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
4074 // CHECK37-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
4075 // CHECK37-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4076 // CHECK37-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
4077 // CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
4078 // CHECK37-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
4079 // CHECK37-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4080 // CHECK37-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
4081 // CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
4082 // CHECK37-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
4083 // CHECK37-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
4084 // CHECK37-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
4085 // CHECK37-NEXT: [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4086 // CHECK37-NEXT: [[TOBOOL2:%.*]] = trunc i8 [[TMP3]] to i1
4087 // CHECK37-NEXT: [[CONV3:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
4088 // CHECK37-NEXT: [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL2]] to i8
4089 // CHECK37-NEXT: store i8 [[FROMBOOL4]], i8* [[CONV3]], align 1
4090 // CHECK37-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4091 // CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
4092 // CHECK37-NEXT: ret void
4093 //
4094 //
4095 // CHECK37-LABEL: define {{[^@]+}}@.omp_outlined.
4096 // CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4097 // CHECK37-NEXT: entry:
4098 // CHECK37-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4099 // CHECK37-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4100 // CHECK37-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
4101 // CHECK37-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
4102 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
4103 // CHECK37-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4104 // CHECK37-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4105 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4
4106 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4107 // CHECK37-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4108 // CHECK37-NEXT: [[I:%.*]] = alloca i32, align 4
4109 // CHECK37-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4110 // CHECK37-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4111 // CHECK37-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4112 // CHECK37-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4113 // CHECK37-NEXT: [[I4:%.*]] = alloca i32, align 4
4114 // CHECK37-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4115 // CHECK37-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4116 // CHECK37-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
4117 // CHECK37-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4118 // CHECK37-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
4119 // CHECK37-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4120 // CHECK37-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
4121 // CHECK37-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4122 // CHECK37-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
4123 // CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
4124 // CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4125 // CHECK37-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4126 // CHECK37-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4127 // CHECK37-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4128 // CHECK37-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4129 // CHECK37-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
4130 // CHECK37-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4131 // CHECK37-NEXT: store i32 0, i32* [[I]], align 4
4132 // CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4133 // CHECK37-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4134 // CHECK37-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4135 // CHECK37: omp.precond.then:
4136 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4137 // CHECK37-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4138 // CHECK37-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
4139 // CHECK37-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4140 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4141 // CHECK37-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4142 // CHECK37-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4143 // CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4144 // CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4145 // CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4146 // CHECK37-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
4147 // CHECK37-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4148 // CHECK37: cond.true:
4149 // CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4150 // CHECK37-NEXT: br label [[COND_END:%.*]]
4151 // CHECK37: cond.false:
4152 // CHECK37-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4153 // CHECK37-NEXT: br label [[COND_END]]
4154 // CHECK37: cond.end:
4155 // CHECK37-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4156 // CHECK37-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4157 // CHECK37-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4158 // CHECK37-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4159 // CHECK37-NEXT: [[TMP14:%.*]] = load i8, i8* [[CONV]], align 1
4160 // CHECK37-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP14]] to i1
4161 // CHECK37-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4162 // CHECK37: omp_if.then:
4163 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4164 // CHECK37: omp.inner.for.cond:
4165 // CHECK37-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
4166 // CHECK37-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
4167 // CHECK37-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
4168 // CHECK37-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4169 // CHECK37: omp.inner.for.body:
4170 // CHECK37-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
4171 // CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
4172 // CHECK37-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4173 // CHECK37-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !5
4174 // CHECK37-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !5
4175 // CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
4176 // CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
4177 // CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !5
4178 // CHECK37-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4179 // CHECK37: omp.body.continue:
4180 // CHECK37-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4181 // CHECK37: omp.inner.for.inc:
4182 // CHECK37-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
4183 // CHECK37-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
4184 // CHECK37-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
4185 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
4186 // CHECK37: omp.inner.for.end:
4187 // CHECK37-NEXT: br label [[OMP_IF_END:%.*]]
4188 // CHECK37: omp_if.else:
4189 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
4190 // CHECK37: omp.inner.for.cond8:
4191 // CHECK37-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4192 // CHECK37-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4193 // CHECK37-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
4194 // CHECK37-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
4195 // CHECK37: omp.inner.for.body10:
4196 // CHECK37-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4197 // CHECK37-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP22]], 1
4198 // CHECK37-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
4199 // CHECK37-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4
4200 // CHECK37-NEXT: [[TMP23:%.*]] = load i32, i32* [[I4]], align 4
4201 // CHECK37-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP23]] to i64
4202 // CHECK37-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM13]]
4203 // CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4
4204 // CHECK37-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
4205 // CHECK37: omp.body.continue15:
4206 // CHECK37-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
4207 // CHECK37: omp.inner.for.inc16:
4208 // CHECK37-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4209 // CHECK37-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP24]], 1
4210 // CHECK37-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
4211 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP9:![0-9]+]]
4212 // CHECK37: omp.inner.for.end18:
4213 // CHECK37-NEXT: br label [[OMP_IF_END]]
4214 // CHECK37: omp_if.end:
4215 // CHECK37-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4216 // CHECK37: omp.loop.exit:
4217 // CHECK37-NEXT: [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4218 // CHECK37-NEXT: [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
4219 // CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
4220 // CHECK37-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4221 // CHECK37-NEXT: [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4222 // CHECK37-NEXT: br i1 [[TMP28]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4223 // CHECK37: .omp.final.then:
4224 // CHECK37-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4225 // CHECK37-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP29]], 0
4226 // CHECK37-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1
4227 // CHECK37-NEXT: [[MUL21:%.*]] = mul nsw i32 [[DIV20]], 1
4228 // CHECK37-NEXT: [[ADD22:%.*]] = add nsw i32 0, [[MUL21]]
4229 // CHECK37-NEXT: store i32 [[ADD22]], i32* [[I4]], align 4
4230 // CHECK37-NEXT: br label [[DOTOMP_FINAL_DONE]]
4231 // CHECK37: .omp.final.done:
4232 // CHECK37-NEXT: br label [[OMP_PRECOND_END]]
4233 // CHECK37: omp.precond.end:
4234 // CHECK37-NEXT: ret void
4235 //
4236 //
4237 // CHECK37-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4238 // CHECK37-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
4239 // CHECK37-NEXT: entry:
4240 // CHECK37-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4241 // CHECK37-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4242 // CHECK37-NEXT: [[TE:%.*]] = alloca i32, align 4
4243 // CHECK37-NEXT: [[TH:%.*]] = alloca i32, align 4
4244 // CHECK37-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
4245 // CHECK37-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
4246 // CHECK37-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
4247 // CHECK37-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
4248 // CHECK37-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
4249 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4
4250 // CHECK37-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4251 // CHECK37-NEXT: store i32 0, i32* [[TE]], align 4
4252 // CHECK37-NEXT: store i32 128, i32* [[TH]], align 4
4253 // CHECK37-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
4254 // CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
4255 // CHECK37-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
4256 // CHECK37-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8
4257 // CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
4258 // CHECK37-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
4259 // CHECK37-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
4260 // CHECK37-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8
4261 // CHECK37-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4262 // CHECK37-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
4263 // CHECK37-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8
4264 // CHECK37-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4265 // CHECK37-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4266 // CHECK37-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8
4267 // CHECK37-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4268 // CHECK37-NEXT: store i8* null, i8** [[TMP8]], align 8
4269 // CHECK37-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4270 // CHECK37-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
4271 // CHECK37-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8
4272 // CHECK37-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4273 // CHECK37-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
4274 // CHECK37-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
4275 // CHECK37-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
4276 // CHECK37-NEXT: store i8* null, i8** [[TMP13]], align 8
4277 // CHECK37-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4278 // CHECK37-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
4279 // CHECK37-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8
4280 // CHECK37-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4281 // CHECK37-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
4282 // CHECK37-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8
4283 // CHECK37-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4284 // CHECK37-NEXT: store i8* null, i8** [[TMP18]], align 8
4285 // CHECK37-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4286 // CHECK37-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4287 // CHECK37-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
4288 // CHECK37-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4289 // CHECK37-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
4290 // CHECK37-NEXT: store i32 1, i32* [[TMP22]], align 4
4291 // CHECK37-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
4292 // CHECK37-NEXT: store i32 3, i32* [[TMP23]], align 4
4293 // CHECK37-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
4294 // CHECK37-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 8
4295 // CHECK37-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
4296 // CHECK37-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 8
4297 // CHECK37-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
4298 // CHECK37-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP26]], align 8
4299 // CHECK37-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
4300 // CHECK37-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP27]], align 8
4301 // CHECK37-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
4302 // CHECK37-NEXT: store i8** null, i8*** [[TMP28]], align 8
4303 // CHECK37-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
4304 // CHECK37-NEXT: store i8** null, i8*** [[TMP29]], align 8
4305 // CHECK37-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
4306 // CHECK37-NEXT: store i64 10, i64* [[TMP30]], align 8
4307 // CHECK37-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
4308 // CHECK37-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
4309 // CHECK37-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4310 // CHECK37: omp_offload.failed:
4311 // CHECK37-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
4312 // CHECK37-NEXT: br label [[OMP_OFFLOAD_CONT]]
4313 // CHECK37: omp_offload.cont:
4314 // CHECK37-NEXT: ret i32 0
4315 //
4316 //
4317 // CHECK37-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181
4318 // CHECK37-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4319 // CHECK37-NEXT: entry:
4320 // CHECK37-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
4321 // CHECK37-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
4322 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4323 // CHECK37-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
4324 // CHECK37-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
4325 // CHECK37-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
4326 // CHECK37-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4327 // CHECK37-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
4328 // CHECK37-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
4329 // CHECK37-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4330 // CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
4331 // CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
4332 // CHECK37-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
4333 // CHECK37-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
4334 // CHECK37-NEXT: ret void
4335 //
4336 //
4337 // CHECK37-LABEL: define {{[^@]+}}@.omp_outlined..1
4338 // CHECK37-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4339 // CHECK37-NEXT: entry:
4340 // CHECK37-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4341 // CHECK37-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4342 // CHECK37-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4343 // CHECK37-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4344 // CHECK37-NEXT: [[TMP:%.*]] = alloca i32, align 4
4345 // CHECK37-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4346 // CHECK37-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4347 // CHECK37-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4348 // CHECK37-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4349 // CHECK37-NEXT: [[I:%.*]] = alloca i32, align 4
4350 // CHECK37-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4351 // CHECK37-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4352 // CHECK37-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4353 // CHECK37-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4354 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4355 // CHECK37-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4356 // CHECK37-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4357 // CHECK37-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4358 // CHECK37-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4359 // CHECK37-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4360 // CHECK37-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4361 // CHECK37-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4362 // CHECK37-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4363 // CHECK37-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4364 // CHECK37: cond.true:
4365 // CHECK37-NEXT: br label [[COND_END:%.*]]
4366 // CHECK37: cond.false:
4367 // CHECK37-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4368 // CHECK37-NEXT: br label [[COND_END]]
4369 // CHECK37: cond.end:
4370 // CHECK37-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4371 // CHECK37-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4372 // CHECK37-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4373 // CHECK37-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4374 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4375 // CHECK37: omp.inner.for.cond:
4376 // CHECK37-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4377 // CHECK37-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
4378 // CHECK37-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4379 // CHECK37-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4380 // CHECK37: omp.inner.for.body:
4381 // CHECK37-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4382 // CHECK37-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4383 // CHECK37-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4384 // CHECK37-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
4385 // CHECK37-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
4386 // CHECK37-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
4387 // CHECK37-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
4388 // CHECK37-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
4389 // CHECK37-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4390 // CHECK37: omp.body.continue:
4391 // CHECK37-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4392 // CHECK37: omp.inner.for.inc:
4393 // CHECK37-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4394 // CHECK37-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
4395 // CHECK37-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4396 // CHECK37-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
4397 // CHECK37: omp.inner.for.end:
4398 // CHECK37-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4399 // CHECK37: omp.loop.exit:
4400 // CHECK37-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4401 // CHECK37-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4402 // CHECK37-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4403 // CHECK37-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4404 // CHECK37: .omp.final.then:
4405 // CHECK37-NEXT: store i32 10, i32* [[I]], align 4
4406 // CHECK37-NEXT: br label [[DOTOMP_FINAL_DONE]]
4407 // CHECK37: .omp.final.done:
4408 // CHECK37-NEXT: ret void
4409 //
4410 //
4411 // CHECK37-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4412 // CHECK37-SAME: () #[[ATTR6:[0-9]+]] {
4413 // CHECK37-NEXT: entry:
4414 // CHECK37-NEXT: call void @__tgt_register_requires(i64 1)
4415 // CHECK37-NEXT: ret void
4416 //
4417 //
4418 // CHECK39-LABEL: define {{[^@]+}}@main
4419 // CHECK39-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4420 // CHECK39-NEXT: entry:
4421 // CHECK39-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
4422 // CHECK39-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4423 // CHECK39-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
4424 // CHECK39-NEXT: [[N:%.*]] = alloca i32, align 4
4425 // CHECK39-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
4426 // CHECK39-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
4427 // CHECK39-NEXT: [[ARGC_CASTED:%.*]] = alloca i32, align 4
4428 // CHECK39-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
4429 // CHECK39-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
4430 // CHECK39-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
4431 // CHECK39-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
4432 // CHECK39-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [4 x i64], align 4
4433 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4
4434 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4435 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4436 // CHECK39-NEXT: store i32 0, i32* [[RETVAL]], align 4
4437 // CHECK39-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4438 // CHECK39-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
4439 // CHECK39-NEXT: store i32 100, i32* [[N]], align 4
4440 // CHECK39-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
4441 // CHECK39-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
4442 // CHECK39-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
4443 // CHECK39-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
4444 // CHECK39-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
4445 // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4446 // CHECK39-NEXT: store i32 [[TMP2]], i32* [[ARGC_CASTED]], align 4
4447 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[ARGC_CASTED]], align 4
4448 // CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[N]], align 4
4449 // CHECK39-NEXT: store i32 [[TMP4]], i32* [[N_CASTED]], align 4
4450 // CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
4451 // CHECK39-NEXT: [[TMP6:%.*]] = mul nuw i32 [[TMP0]], 4
4452 // CHECK39-NEXT: [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
4453 // CHECK39-NEXT: [[TMP8:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
4454 // CHECK39-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP8]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes to i8*), i32 32, i1 false)
4455 // CHECK39-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4456 // CHECK39-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4457 // CHECK39-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
4458 // CHECK39-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4459 // CHECK39-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4460 // CHECK39-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
4461 // CHECK39-NEXT: [[TMP13:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4462 // CHECK39-NEXT: store i8* null, i8** [[TMP13]], align 4
4463 // CHECK39-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4464 // CHECK39-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4465 // CHECK39-NEXT: store i32 [[TMP5]], i32* [[TMP15]], align 4
4466 // CHECK39-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4467 // CHECK39-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to i32*
4468 // CHECK39-NEXT: store i32 [[TMP5]], i32* [[TMP17]], align 4
4469 // CHECK39-NEXT: [[TMP18:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4470 // CHECK39-NEXT: store i8* null, i8** [[TMP18]], align 4
4471 // CHECK39-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4472 // CHECK39-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
4473 // CHECK39-NEXT: store i32 [[TMP0]], i32* [[TMP20]], align 4
4474 // CHECK39-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4475 // CHECK39-NEXT: [[TMP22:%.*]] = bitcast i8** [[TMP21]] to i32*
4476 // CHECK39-NEXT: store i32 [[TMP0]], i32* [[TMP22]], align 4
4477 // CHECK39-NEXT: [[TMP23:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4478 // CHECK39-NEXT: store i8* null, i8** [[TMP23]], align 4
4479 // CHECK39-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
4480 // CHECK39-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32**
4481 // CHECK39-NEXT: store i32* [[VLA]], i32** [[TMP25]], align 4
4482 // CHECK39-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
4483 // CHECK39-NEXT: [[TMP27:%.*]] = bitcast i8** [[TMP26]] to i32**
4484 // CHECK39-NEXT: store i32* [[VLA]], i32** [[TMP27]], align 4
4485 // CHECK39-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 3
4486 // CHECK39-NEXT: store i64 [[TMP7]], i64* [[TMP28]], align 4
4487 // CHECK39-NEXT: [[TMP29:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
4488 // CHECK39-NEXT: store i8* null, i8** [[TMP29]], align 4
4489 // CHECK39-NEXT: [[TMP30:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4490 // CHECK39-NEXT: [[TMP31:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4491 // CHECK39-NEXT: [[TMP32:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4492 // CHECK39-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
4493 // CHECK39-NEXT: store i32 [[TMP33]], i32* [[DOTCAPTURE_EXPR_]], align 4
4494 // CHECK39-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4495 // CHECK39-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP34]], 0
4496 // CHECK39-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4497 // CHECK39-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4498 // CHECK39-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4499 // CHECK39-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4500 // CHECK39-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP35]], 1
4501 // CHECK39-NEXT: [[TMP36:%.*]] = zext i32 [[ADD]] to i64
4502 // CHECK39-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4503 // CHECK39-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
4504 // CHECK39-NEXT: store i32 1, i32* [[TMP37]], align 4
4505 // CHECK39-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
4506 // CHECK39-NEXT: store i32 4, i32* [[TMP38]], align 4
4507 // CHECK39-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
4508 // CHECK39-NEXT: store i8** [[TMP30]], i8*** [[TMP39]], align 4
4509 // CHECK39-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
4510 // CHECK39-NEXT: store i8** [[TMP31]], i8*** [[TMP40]], align 4
4511 // CHECK39-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
4512 // CHECK39-NEXT: store i64* [[TMP32]], i64** [[TMP41]], align 4
4513 // CHECK39-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
4514 // CHECK39-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP42]], align 4
4515 // CHECK39-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
4516 // CHECK39-NEXT: store i8** null, i8*** [[TMP43]], align 4
4517 // CHECK39-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
4518 // CHECK39-NEXT: store i8** null, i8*** [[TMP44]], align 4
4519 // CHECK39-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
4520 // CHECK39-NEXT: store i64 [[TMP36]], i64* [[TMP45]], align 8
4521 // CHECK39-NEXT: [[TMP46:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
4522 // CHECK39-NEXT: [[TMP47:%.*]] = icmp ne i32 [[TMP46]], 0
4523 // CHECK39-NEXT: br i1 [[TMP47]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4524 // CHECK39: omp_offload.failed:
4525 // CHECK39-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192(i32 [[TMP3]], i32 [[TMP5]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
4526 // CHECK39-NEXT: br label [[OMP_OFFLOAD_CONT]]
4527 // CHECK39: omp_offload.cont:
4528 // CHECK39-NEXT: [[TMP48:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4529 // CHECK39-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP48]])
4530 // CHECK39-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
4531 // CHECK39-NEXT: [[TMP49:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4532 // CHECK39-NEXT: call void @llvm.stackrestore(i8* [[TMP49]])
4533 // CHECK39-NEXT: [[TMP50:%.*]] = load i32, i32* [[RETVAL]], align 4
4534 // CHECK39-NEXT: ret i32 [[TMP50]]
4535 //
4536 //
4537 // CHECK39-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l192
4538 // CHECK39-SAME: (i32 noundef [[ARGC:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
4539 // CHECK39-NEXT: entry:
4540 // CHECK39-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4541 // CHECK39-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
4542 // CHECK39-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4543 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4544 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4545 // CHECK39-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4546 // CHECK39-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4547 // CHECK39-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
4548 // CHECK39-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4549 // CHECK39-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4550 // CHECK39-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4551 // CHECK39-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4552 // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4553 // CHECK39-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
4554 // CHECK39-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
4555 // CHECK39-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
4556 // CHECK39-NEXT: [[TMP3:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4557 // CHECK39-NEXT: [[TOBOOL1:%.*]] = trunc i8 [[TMP3]] to i1
4558 // CHECK39-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__CASTED]] to i8*
4559 // CHECK39-NEXT: [[FROMBOOL2:%.*]] = zext i1 [[TOBOOL1]] to i8
4560 // CHECK39-NEXT: store i8 [[FROMBOOL2]], i8* [[CONV]], align 1
4561 // CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4562 // CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
4563 // CHECK39-NEXT: ret void
4564 //
4565 //
4566 // CHECK39-LABEL: define {{[^@]+}}@.omp_outlined.
4567 // CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4568 // CHECK39-NEXT: entry:
4569 // CHECK39-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4570 // CHECK39-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4571 // CHECK39-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
4572 // CHECK39-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4573 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4574 // CHECK39-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4575 // CHECK39-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4576 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4
4577 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4578 // CHECK39-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4579 // CHECK39-NEXT: [[I:%.*]] = alloca i32, align 4
4580 // CHECK39-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4581 // CHECK39-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4582 // CHECK39-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4583 // CHECK39-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4584 // CHECK39-NEXT: [[I4:%.*]] = alloca i32, align 4
4585 // CHECK39-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4586 // CHECK39-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4587 // CHECK39-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
4588 // CHECK39-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4589 // CHECK39-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4590 // CHECK39-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4591 // CHECK39-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
4592 // CHECK39-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4593 // CHECK39-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4594 // CHECK39-NEXT: [[CONV:%.*]] = bitcast i32* [[DOTCAPTURE_EXPR__ADDR]] to i8*
4595 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4596 // CHECK39-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4597 // CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4598 // CHECK39-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4599 // CHECK39-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4600 // CHECK39-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
4601 // CHECK39-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4602 // CHECK39-NEXT: store i32 0, i32* [[I]], align 4
4603 // CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4604 // CHECK39-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4605 // CHECK39-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4606 // CHECK39: omp.precond.then:
4607 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4608 // CHECK39-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4609 // CHECK39-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
4610 // CHECK39-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4611 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4612 // CHECK39-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4613 // CHECK39-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4614 // CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4615 // CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4616 // CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4617 // CHECK39-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
4618 // CHECK39-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4619 // CHECK39: cond.true:
4620 // CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4621 // CHECK39-NEXT: br label [[COND_END:%.*]]
4622 // CHECK39: cond.false:
4623 // CHECK39-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4624 // CHECK39-NEXT: br label [[COND_END]]
4625 // CHECK39: cond.end:
4626 // CHECK39-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4627 // CHECK39-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4628 // CHECK39-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4629 // CHECK39-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4630 // CHECK39-NEXT: [[TMP14:%.*]] = load i8, i8* [[CONV]], align 1
4631 // CHECK39-NEXT: [[TOBOOL:%.*]] = trunc i8 [[TMP14]] to i1
4632 // CHECK39-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4633 // CHECK39: omp_if.then:
4634 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4635 // CHECK39: omp.inner.for.cond:
4636 // CHECK39-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4637 // CHECK39-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
4638 // CHECK39-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
4639 // CHECK39-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4640 // CHECK39: omp.inner.for.body:
4641 // CHECK39-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4642 // CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP17]], 1
4643 // CHECK39-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4644 // CHECK39-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !6
4645 // CHECK39-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !6
4646 // CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP18]]
4647 // CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6
4648 // CHECK39-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4649 // CHECK39: omp.body.continue:
4650 // CHECK39-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4651 // CHECK39: omp.inner.for.inc:
4652 // CHECK39-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4653 // CHECK39-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP19]], 1
4654 // CHECK39-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4655 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
4656 // CHECK39: omp.inner.for.end:
4657 // CHECK39-NEXT: br label [[OMP_IF_END:%.*]]
4658 // CHECK39: omp_if.else:
4659 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
4660 // CHECK39: omp.inner.for.cond8:
4661 // CHECK39-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4662 // CHECK39-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4663 // CHECK39-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
4664 // CHECK39-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
4665 // CHECK39: omp.inner.for.body10:
4666 // CHECK39-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4667 // CHECK39-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP22]], 1
4668 // CHECK39-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
4669 // CHECK39-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4
4670 // CHECK39-NEXT: [[TMP23:%.*]] = load i32, i32* [[I4]], align 4
4671 // CHECK39-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP23]]
4672 // CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4
4673 // CHECK39-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
4674 // CHECK39: omp.body.continue14:
4675 // CHECK39-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
4676 // CHECK39: omp.inner.for.inc15:
4677 // CHECK39-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4678 // CHECK39-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP24]], 1
4679 // CHECK39-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4
4680 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP10:![0-9]+]]
4681 // CHECK39: omp.inner.for.end17:
4682 // CHECK39-NEXT: br label [[OMP_IF_END]]
4683 // CHECK39: omp_if.end:
4684 // CHECK39-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4685 // CHECK39: omp.loop.exit:
4686 // CHECK39-NEXT: [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4687 // CHECK39-NEXT: [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
4688 // CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
4689 // CHECK39-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4690 // CHECK39-NEXT: [[TMP28:%.*]] = icmp ne i32 [[TMP27]], 0
4691 // CHECK39-NEXT: br i1 [[TMP28]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4692 // CHECK39: .omp.final.then:
4693 // CHECK39-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4694 // CHECK39-NEXT: [[SUB18:%.*]] = sub nsw i32 [[TMP29]], 0
4695 // CHECK39-NEXT: [[DIV19:%.*]] = sdiv i32 [[SUB18]], 1
4696 // CHECK39-NEXT: [[MUL20:%.*]] = mul nsw i32 [[DIV19]], 1
4697 // CHECK39-NEXT: [[ADD21:%.*]] = add nsw i32 0, [[MUL20]]
4698 // CHECK39-NEXT: store i32 [[ADD21]], i32* [[I4]], align 4
4699 // CHECK39-NEXT: br label [[DOTOMP_FINAL_DONE]]
4700 // CHECK39: .omp.final.done:
4701 // CHECK39-NEXT: br label [[OMP_PRECOND_END]]
4702 // CHECK39: omp.precond.end:
4703 // CHECK39-NEXT: ret void
4704 //
4705 //
4706 // CHECK39-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4707 // CHECK39-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
4708 // CHECK39-NEXT: entry:
4709 // CHECK39-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4710 // CHECK39-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4711 // CHECK39-NEXT: [[TE:%.*]] = alloca i32, align 4
4712 // CHECK39-NEXT: [[TH:%.*]] = alloca i32, align 4
4713 // CHECK39-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
4714 // CHECK39-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
4715 // CHECK39-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4716 // CHECK39-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4717 // CHECK39-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4718 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4
4719 // CHECK39-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4720 // CHECK39-NEXT: store i32 0, i32* [[TE]], align 4
4721 // CHECK39-NEXT: store i32 128, i32* [[TH]], align 4
4722 // CHECK39-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
4723 // CHECK39-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4
4724 // CHECK39-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4
4725 // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
4726 // CHECK39-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4
4727 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4
4728 // CHECK39-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4729 // CHECK39-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
4730 // CHECK39-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4
4731 // CHECK39-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4732 // CHECK39-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
4733 // CHECK39-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4
4734 // CHECK39-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4735 // CHECK39-NEXT: store i8* null, i8** [[TMP8]], align 4
4736 // CHECK39-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4737 // CHECK39-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4738 // CHECK39-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
4739 // CHECK39-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4740 // CHECK39-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4741 // CHECK39-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
4742 // CHECK39-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4743 // CHECK39-NEXT: store i8* null, i8** [[TMP13]], align 4
4744 // CHECK39-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4745 // CHECK39-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
4746 // CHECK39-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4
4747 // CHECK39-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4748 // CHECK39-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
4749 // CHECK39-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4
4750 // CHECK39-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4751 // CHECK39-NEXT: store i8* null, i8** [[TMP18]], align 4
4752 // CHECK39-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4753 // CHECK39-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4754 // CHECK39-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
4755 // CHECK39-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4756 // CHECK39-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
4757 // CHECK39-NEXT: store i32 1, i32* [[TMP22]], align 4
4758 // CHECK39-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
4759 // CHECK39-NEXT: store i32 3, i32* [[TMP23]], align 4
4760 // CHECK39-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
4761 // CHECK39-NEXT: store i8** [[TMP19]], i8*** [[TMP24]], align 4
4762 // CHECK39-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
4763 // CHECK39-NEXT: store i8** [[TMP20]], i8*** [[TMP25]], align 4
4764 // CHECK39-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
4765 // CHECK39-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP26]], align 4
4766 // CHECK39-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
4767 // CHECK39-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP27]], align 4
4768 // CHECK39-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
4769 // CHECK39-NEXT: store i8** null, i8*** [[TMP28]], align 4
4770 // CHECK39-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
4771 // CHECK39-NEXT: store i8** null, i8*** [[TMP29]], align 4
4772 // CHECK39-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
4773 // CHECK39-NEXT: store i64 10, i64* [[TMP30]], align 8
4774 // CHECK39-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
4775 // CHECK39-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
4776 // CHECK39-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4777 // CHECK39: omp_offload.failed:
4778 // CHECK39-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
4779 // CHECK39-NEXT: br label [[OMP_OFFLOAD_CONT]]
4780 // CHECK39: omp_offload.cont:
4781 // CHECK39-NEXT: ret i32 0
4782 //
4783 //
4784 // CHECK39-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l181
4785 // CHECK39-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4786 // CHECK39-NEXT: entry:
4787 // CHECK39-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
4788 // CHECK39-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
4789 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4790 // CHECK39-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
4791 // CHECK39-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
4792 // CHECK39-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
4793 // CHECK39-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4794 // CHECK39-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4795 // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
4796 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
4797 // CHECK39-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
4798 // CHECK39-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
4799 // CHECK39-NEXT: ret void
4800 //
4801 //
4802 // CHECK39-LABEL: define {{[^@]+}}@.omp_outlined..1
4803 // CHECK39-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4804 // CHECK39-NEXT: entry:
4805 // CHECK39-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4806 // CHECK39-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4807 // CHECK39-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4808 // CHECK39-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4809 // CHECK39-NEXT: [[TMP:%.*]] = alloca i32, align 4
4810 // CHECK39-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4811 // CHECK39-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4812 // CHECK39-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4813 // CHECK39-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4814 // CHECK39-NEXT: [[I:%.*]] = alloca i32, align 4
4815 // CHECK39-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4816 // CHECK39-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4817 // CHECK39-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4818 // CHECK39-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4819 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4820 // CHECK39-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4821 // CHECK39-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4822 // CHECK39-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4823 // CHECK39-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4824 // CHECK39-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4825 // CHECK39-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4826 // CHECK39-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4827 // CHECK39-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4828 // CHECK39-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4829 // CHECK39: cond.true:
4830 // CHECK39-NEXT: br label [[COND_END:%.*]]
4831 // CHECK39: cond.false:
4832 // CHECK39-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4833 // CHECK39-NEXT: br label [[COND_END]]
4834 // CHECK39: cond.end:
4835 // CHECK39-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4836 // CHECK39-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4837 // CHECK39-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4838 // CHECK39-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4839 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4840 // CHECK39: omp.inner.for.cond:
4841 // CHECK39-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4842 // CHECK39-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
4843 // CHECK39-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4844 // CHECK39-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4845 // CHECK39: omp.inner.for.body:
4846 // CHECK39-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4847 // CHECK39-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4848 // CHECK39-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4849 // CHECK39-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
4850 // CHECK39-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
4851 // CHECK39-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
4852 // CHECK39-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
4853 // CHECK39-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4854 // CHECK39: omp.body.continue:
4855 // CHECK39-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4856 // CHECK39: omp.inner.for.inc:
4857 // CHECK39-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4858 // CHECK39-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
4859 // CHECK39-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4860 // CHECK39-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
4861 // CHECK39: omp.inner.for.end:
4862 // CHECK39-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4863 // CHECK39: omp.loop.exit:
4864 // CHECK39-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4865 // CHECK39-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4866 // CHECK39-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4867 // CHECK39-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4868 // CHECK39: .omp.final.then:
4869 // CHECK39-NEXT: store i32 10, i32* [[I]], align 4
4870 // CHECK39-NEXT: br label [[DOTOMP_FINAL_DONE]]
4871 // CHECK39: .omp.final.done:
4872 // CHECK39-NEXT: ret void
4873 //
4874 //
4875 // CHECK39-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4876 // CHECK39-SAME: () #[[ATTR6:[0-9]+]] {
4877 // CHECK39-NEXT: entry:
4878 // CHECK39-NEXT: call void @__tgt_register_requires(i64 1)
4879 // CHECK39-NEXT: ret void
4880 //
4881 //
4882 // CHECK41-LABEL: define {{[^@]+}}@main
4883 // CHECK41-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4884 // CHECK41-NEXT: entry:
4885 // CHECK41-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
4886 // CHECK41-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4887 // CHECK41-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
4888 // CHECK41-NEXT: [[N:%.*]] = alloca i32, align 4
4889 // CHECK41-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
4890 // CHECK41-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
4891 // CHECK41-NEXT: [[TMP:%.*]] = alloca i32, align 4
4892 // CHECK41-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4893 // CHECK41-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4894 // CHECK41-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4895 // CHECK41-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4896 // CHECK41-NEXT: [[I:%.*]] = alloca i32, align 4
4897 // CHECK41-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4898 // CHECK41-NEXT: [[I3:%.*]] = alloca i32, align 4
4899 // CHECK41-NEXT: store i32 0, i32* [[RETVAL]], align 4
4900 // CHECK41-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4901 // CHECK41-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
4902 // CHECK41-NEXT: store i32 100, i32* [[N]], align 4
4903 // CHECK41-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
4904 // CHECK41-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
4905 // CHECK41-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
4906 // CHECK41-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
4907 // CHECK41-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
4908 // CHECK41-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
4909 // CHECK41-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
4910 // CHECK41-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
4911 // CHECK41-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4912 // CHECK41-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4913 // CHECK41-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4914 // CHECK41-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4915 // CHECK41-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4916 // CHECK41-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4917 // CHECK41-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4918 // CHECK41-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
4919 // CHECK41-NEXT: store i32 0, i32* [[I]], align 4
4920 // CHECK41-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4921 // CHECK41-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
4922 // CHECK41-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
4923 // CHECK41: simd.if.then:
4924 // CHECK41-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4925 // CHECK41-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4926 // CHECK41-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4927 // CHECK41: omp.inner.for.cond:
4928 // CHECK41-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
4929 // CHECK41-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
4930 // CHECK41-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4931 // CHECK41-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4932 // CHECK41: omp.inner.for.body:
4933 // CHECK41-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
4934 // CHECK41-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4935 // CHECK41-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4936 // CHECK41-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
4937 // CHECK41-NEXT: [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
4938 // CHECK41-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4939 // CHECK41-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
4940 // CHECK41-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
4941 // CHECK41-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4942 // CHECK41: omp.body.continue:
4943 // CHECK41-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4944 // CHECK41: omp.inner.for.inc:
4945 // CHECK41-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
4946 // CHECK41-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
4947 // CHECK41-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
4948 // CHECK41-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
4949 // CHECK41: omp.inner.for.end:
4950 // CHECK41-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4951 // CHECK41-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
4952 // CHECK41-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
4953 // CHECK41-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
4954 // CHECK41-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
4955 // CHECK41-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
4956 // CHECK41-NEXT: br label [[SIMD_IF_END]]
4957 // CHECK41: simd.if.end:
4958 // CHECK41-NEXT: [[TMP14:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4959 // CHECK41-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP14]])
4960 // CHECK41-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
4961 // CHECK41-NEXT: [[TMP15:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4962 // CHECK41-NEXT: call void @llvm.stackrestore(i8* [[TMP15]])
4963 // CHECK41-NEXT: [[TMP16:%.*]] = load i32, i32* [[RETVAL]], align 4
4964 // CHECK41-NEXT: ret i32 [[TMP16]]
4965 //
4966 //
4967 // CHECK41-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4968 // CHECK41-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
4969 // CHECK41-NEXT: entry:
4970 // CHECK41-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4971 // CHECK41-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4972 // CHECK41-NEXT: [[TE:%.*]] = alloca i32, align 4
4973 // CHECK41-NEXT: [[TH:%.*]] = alloca i32, align 4
4974 // CHECK41-NEXT: [[TMP:%.*]] = alloca i32, align 4
4975 // CHECK41-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4976 // CHECK41-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4977 // CHECK41-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4978 // CHECK41-NEXT: [[I:%.*]] = alloca i32, align 4
4979 // CHECK41-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4980 // CHECK41-NEXT: store i32 0, i32* [[TE]], align 4
4981 // CHECK41-NEXT: store i32 128, i32* [[TH]], align 4
4982 // CHECK41-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4983 // CHECK41-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4984 // CHECK41-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4985 // CHECK41-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
4986 // CHECK41-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4987 // CHECK41: omp.inner.for.cond:
4988 // CHECK41-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4989 // CHECK41-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
4990 // CHECK41-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
4991 // CHECK41-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4992 // CHECK41: omp.inner.for.body:
4993 // CHECK41-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
4994 // CHECK41-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
4995 // CHECK41-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4996 // CHECK41-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
4997 // CHECK41-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !6
4998 // CHECK41-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
4999 // CHECK41-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5000 // CHECK41-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !6
5001 // CHECK41-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5002 // CHECK41: omp.body.continue:
5003 // CHECK41-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5004 // CHECK41: omp.inner.for.inc:
5005 // CHECK41-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
5006 // CHECK41-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5007 // CHECK41-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
5008 // CHECK41-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
5009 // CHECK41: omp.inner.for.end:
5010 // CHECK41-NEXT: store i32 10, i32* [[I]], align 4
5011 // CHECK41-NEXT: ret i32 0
5012 //
5013 //
5014 // CHECK43-LABEL: define {{[^@]+}}@main
5015 // CHECK43-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
5016 // CHECK43-NEXT: entry:
5017 // CHECK43-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5018 // CHECK43-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5019 // CHECK43-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
5020 // CHECK43-NEXT: [[N:%.*]] = alloca i32, align 4
5021 // CHECK43-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
5022 // CHECK43-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5023 // CHECK43-NEXT: [[TMP:%.*]] = alloca i32, align 4
5024 // CHECK43-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5025 // CHECK43-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5026 // CHECK43-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5027 // CHECK43-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5028 // CHECK43-NEXT: [[I:%.*]] = alloca i32, align 4
5029 // CHECK43-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5030 // CHECK43-NEXT: [[I3:%.*]] = alloca i32, align 4
5031 // CHECK43-NEXT: store i32 0, i32* [[RETVAL]], align 4
5032 // CHECK43-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5033 // CHECK43-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
5034 // CHECK43-NEXT: store i32 100, i32* [[N]], align 4
5035 // CHECK43-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
5036 // CHECK43-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
5037 // CHECK43-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
5038 // CHECK43-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
5039 // CHECK43-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
5040 // CHECK43-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
5041 // CHECK43-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
5042 // CHECK43-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5043 // CHECK43-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
5044 // CHECK43-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5045 // CHECK43-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5046 // CHECK43-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5047 // CHECK43-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5048 // CHECK43-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5049 // CHECK43-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
5050 // CHECK43-NEXT: store i32 0, i32* [[I]], align 4
5051 // CHECK43-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5052 // CHECK43-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5053 // CHECK43-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
5054 // CHECK43: simd.if.then:
5055 // CHECK43-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5056 // CHECK43-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5057 // CHECK43-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5058 // CHECK43: omp.inner.for.cond:
5059 // CHECK43-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5060 // CHECK43-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
5061 // CHECK43-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5062 // CHECK43-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5063 // CHECK43: omp.inner.for.body:
5064 // CHECK43-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5065 // CHECK43-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5066 // CHECK43-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5067 // CHECK43-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
5068 // CHECK43-NEXT: [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
5069 // CHECK43-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
5070 // CHECK43-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
5071 // CHECK43-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5072 // CHECK43: omp.body.continue:
5073 // CHECK43-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5074 // CHECK43: omp.inner.for.inc:
5075 // CHECK43-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5076 // CHECK43-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
5077 // CHECK43-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5078 // CHECK43-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
5079 // CHECK43: omp.inner.for.end:
5080 // CHECK43-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5081 // CHECK43-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
5082 // CHECK43-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
5083 // CHECK43-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
5084 // CHECK43-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
5085 // CHECK43-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
5086 // CHECK43-NEXT: br label [[SIMD_IF_END]]
5087 // CHECK43: simd.if.end:
5088 // CHECK43-NEXT: [[TMP13:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5089 // CHECK43-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP13]])
5090 // CHECK43-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
5091 // CHECK43-NEXT: [[TMP14:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5092 // CHECK43-NEXT: call void @llvm.stackrestore(i8* [[TMP14]])
5093 // CHECK43-NEXT: [[TMP15:%.*]] = load i32, i32* [[RETVAL]], align 4
5094 // CHECK43-NEXT: ret i32 [[TMP15]]
5095 //
5096 //
5097 // CHECK43-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5098 // CHECK43-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
5099 // CHECK43-NEXT: entry:
5100 // CHECK43-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5101 // CHECK43-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5102 // CHECK43-NEXT: [[TE:%.*]] = alloca i32, align 4
5103 // CHECK43-NEXT: [[TH:%.*]] = alloca i32, align 4
5104 // CHECK43-NEXT: [[TMP:%.*]] = alloca i32, align 4
5105 // CHECK43-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5106 // CHECK43-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5107 // CHECK43-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5108 // CHECK43-NEXT: [[I:%.*]] = alloca i32, align 4
5109 // CHECK43-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5110 // CHECK43-NEXT: store i32 0, i32* [[TE]], align 4
5111 // CHECK43-NEXT: store i32 128, i32* [[TH]], align 4
5112 // CHECK43-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5113 // CHECK43-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5114 // CHECK43-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5115 // CHECK43-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
5116 // CHECK43-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5117 // CHECK43: omp.inner.for.cond:
5118 // CHECK43-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
5119 // CHECK43-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
5120 // CHECK43-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
5121 // CHECK43-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5122 // CHECK43: omp.inner.for.body:
5123 // CHECK43-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
5124 // CHECK43-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
5125 // CHECK43-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5126 // CHECK43-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !7
5127 // CHECK43-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !7
5128 // CHECK43-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
5129 // CHECK43-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !7
5130 // CHECK43-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5131 // CHECK43: omp.body.continue:
5132 // CHECK43-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5133 // CHECK43: omp.inner.for.inc:
5134 // CHECK43-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
5135 // CHECK43-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5136 // CHECK43-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
5137 // CHECK43-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
5138 // CHECK43: omp.inner.for.end:
5139 // CHECK43-NEXT: store i32 10, i32* [[I]], align 4
5140 // CHECK43-NEXT: ret i32 0
5141 //
5142 //
5143 // CHECK45-LABEL: define {{[^@]+}}@main
5144 // CHECK45-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
5145 // CHECK45-NEXT: entry:
5146 // CHECK45-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5147 // CHECK45-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5148 // CHECK45-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
5149 // CHECK45-NEXT: [[N:%.*]] = alloca i32, align 4
5150 // CHECK45-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
5151 // CHECK45-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
5152 // CHECK45-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
5153 // CHECK45-NEXT: [[TMP:%.*]] = alloca i32, align 4
5154 // CHECK45-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5155 // CHECK45-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5156 // CHECK45-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5157 // CHECK45-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5158 // CHECK45-NEXT: [[I:%.*]] = alloca i32, align 4
5159 // CHECK45-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5160 // CHECK45-NEXT: [[I4:%.*]] = alloca i32, align 4
5161 // CHECK45-NEXT: store i32 0, i32* [[RETVAL]], align 4
5162 // CHECK45-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5163 // CHECK45-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
5164 // CHECK45-NEXT: store i32 100, i32* [[N]], align 4
5165 // CHECK45-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
5166 // CHECK45-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
5167 // CHECK45-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
5168 // CHECK45-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
5169 // CHECK45-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
5170 // CHECK45-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
5171 // CHECK45-NEXT: [[TMP3:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5172 // CHECK45-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP3]], 0
5173 // CHECK45-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
5174 // CHECK45-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
5175 // CHECK45-NEXT: [[TMP4:%.*]] = load i32, i32* [[N]], align 4
5176 // CHECK45-NEXT: store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5177 // CHECK45-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5178 // CHECK45-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
5179 // CHECK45-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5180 // CHECK45-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
5181 // CHECK45-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5182 // CHECK45-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5183 // CHECK45-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5184 // CHECK45-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
5185 // CHECK45-NEXT: store i32 0, i32* [[I]], align 4
5186 // CHECK45-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5187 // CHECK45-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
5188 // CHECK45-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
5189 // CHECK45: simd.if.then:
5190 // CHECK45-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5191 // CHECK45-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
5192 // CHECK45-NEXT: [[TMP9:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5193 // CHECK45-NEXT: [[TOBOOL5:%.*]] = trunc i8 [[TMP9]] to i1
5194 // CHECK45-NEXT: br i1 [[TOBOOL5]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5195 // CHECK45: omp_if.then:
5196 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5197 // CHECK45: omp.inner.for.cond:
5198 // CHECK45-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5199 // CHECK45-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
5200 // CHECK45-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
5201 // CHECK45-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5202 // CHECK45: omp.inner.for.body:
5203 // CHECK45-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5204 // CHECK45-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
5205 // CHECK45-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5206 // CHECK45-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !2
5207 // CHECK45-NEXT: [[TMP13:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !2
5208 // CHECK45-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
5209 // CHECK45-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
5210 // CHECK45-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
5211 // CHECK45-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5212 // CHECK45: omp.body.continue:
5213 // CHECK45-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5214 // CHECK45: omp.inner.for.inc:
5215 // CHECK45-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5216 // CHECK45-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP14]], 1
5217 // CHECK45-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5218 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
5219 // CHECK45: omp.inner.for.end:
5220 // CHECK45-NEXT: br label [[OMP_IF_END:%.*]]
5221 // CHECK45: omp_if.else:
5222 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
5223 // CHECK45: omp.inner.for.cond8:
5224 // CHECK45-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5225 // CHECK45-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5226 // CHECK45-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
5227 // CHECK45-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
5228 // CHECK45: omp.inner.for.body10:
5229 // CHECK45-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5230 // CHECK45-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP17]], 1
5231 // CHECK45-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
5232 // CHECK45-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4
5233 // CHECK45-NEXT: [[TMP18:%.*]] = load i32, i32* [[I4]], align 4
5234 // CHECK45-NEXT: [[IDXPROM13:%.*]] = sext i32 [[TMP18]] to i64
5235 // CHECK45-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM13]]
5236 // CHECK45-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4
5237 // CHECK45-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
5238 // CHECK45: omp.body.continue15:
5239 // CHECK45-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
5240 // CHECK45: omp.inner.for.inc16:
5241 // CHECK45-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5242 // CHECK45-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP19]], 1
5243 // CHECK45-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
5244 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP6:![0-9]+]]
5245 // CHECK45: omp.inner.for.end18:
5246 // CHECK45-NEXT: br label [[OMP_IF_END]]
5247 // CHECK45: omp_if.end:
5248 // CHECK45-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5249 // CHECK45-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP20]], 0
5250 // CHECK45-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1
5251 // CHECK45-NEXT: [[MUL21:%.*]] = mul nsw i32 [[DIV20]], 1
5252 // CHECK45-NEXT: [[ADD22:%.*]] = add nsw i32 0, [[MUL21]]
5253 // CHECK45-NEXT: store i32 [[ADD22]], i32* [[I4]], align 4
5254 // CHECK45-NEXT: br label [[SIMD_IF_END]]
5255 // CHECK45: simd.if.end:
5256 // CHECK45-NEXT: [[TMP21:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5257 // CHECK45-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP21]])
5258 // CHECK45-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
5259 // CHECK45-NEXT: [[TMP22:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
5260 // CHECK45-NEXT: call void @llvm.stackrestore(i8* [[TMP22]])
5261 // CHECK45-NEXT: [[TMP23:%.*]] = load i32, i32* [[RETVAL]], align 4
5262 // CHECK45-NEXT: ret i32 [[TMP23]]
5263 //
5264 //
5265 // CHECK45-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5266 // CHECK45-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
5267 // CHECK45-NEXT: entry:
5268 // CHECK45-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5269 // CHECK45-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5270 // CHECK45-NEXT: [[TE:%.*]] = alloca i32, align 4
5271 // CHECK45-NEXT: [[TH:%.*]] = alloca i32, align 4
5272 // CHECK45-NEXT: [[TMP:%.*]] = alloca i32, align 4
5273 // CHECK45-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5274 // CHECK45-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5275 // CHECK45-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5276 // CHECK45-NEXT: [[I:%.*]] = alloca i32, align 4
5277 // CHECK45-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5278 // CHECK45-NEXT: store i32 0, i32* [[TE]], align 4
5279 // CHECK45-NEXT: store i32 128, i32* [[TH]], align 4
5280 // CHECK45-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5281 // CHECK45-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5282 // CHECK45-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5283 // CHECK45-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
5284 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5285 // CHECK45: omp.inner.for.cond:
5286 // CHECK45-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
5287 // CHECK45-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !8
5288 // CHECK45-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
5289 // CHECK45-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5290 // CHECK45: omp.inner.for.body:
5291 // CHECK45-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
5292 // CHECK45-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
5293 // CHECK45-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5294 // CHECK45-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !8
5295 // CHECK45-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !8
5296 // CHECK45-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
5297 // CHECK45-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5298 // CHECK45-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !8
5299 // CHECK45-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5300 // CHECK45: omp.body.continue:
5301 // CHECK45-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5302 // CHECK45: omp.inner.for.inc:
5303 // CHECK45-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
5304 // CHECK45-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5305 // CHECK45-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
5306 // CHECK45-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
5307 // CHECK45: omp.inner.for.end:
5308 // CHECK45-NEXT: store i32 10, i32* [[I]], align 4
5309 // CHECK45-NEXT: ret i32 0
5310 //
5311 //
5312 // CHECK47-LABEL: define {{[^@]+}}@main
5313 // CHECK47-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
5314 // CHECK47-NEXT: entry:
5315 // CHECK47-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5316 // CHECK47-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5317 // CHECK47-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
5318 // CHECK47-NEXT: [[N:%.*]] = alloca i32, align 4
5319 // CHECK47-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
5320 // CHECK47-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5321 // CHECK47-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
5322 // CHECK47-NEXT: [[TMP:%.*]] = alloca i32, align 4
5323 // CHECK47-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5324 // CHECK47-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5325 // CHECK47-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5326 // CHECK47-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5327 // CHECK47-NEXT: [[I:%.*]] = alloca i32, align 4
5328 // CHECK47-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5329 // CHECK47-NEXT: [[I4:%.*]] = alloca i32, align 4
5330 // CHECK47-NEXT: store i32 0, i32* [[RETVAL]], align 4
5331 // CHECK47-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5332 // CHECK47-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
5333 // CHECK47-NEXT: store i32 100, i32* [[N]], align 4
5334 // CHECK47-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
5335 // CHECK47-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
5336 // CHECK47-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
5337 // CHECK47-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
5338 // CHECK47-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
5339 // CHECK47-NEXT: [[TMP2:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5340 // CHECK47-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP2]], 0
5341 // CHECK47-NEXT: [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
5342 // CHECK47-NEXT: store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
5343 // CHECK47-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
5344 // CHECK47-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5345 // CHECK47-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5346 // CHECK47-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5347 // CHECK47-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5348 // CHECK47-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
5349 // CHECK47-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5350 // CHECK47-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5351 // CHECK47-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5352 // CHECK47-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
5353 // CHECK47-NEXT: store i32 0, i32* [[I]], align 4
5354 // CHECK47-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5355 // CHECK47-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
5356 // CHECK47-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
5357 // CHECK47: simd.if.then:
5358 // CHECK47-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5359 // CHECK47-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5360 // CHECK47-NEXT: [[TMP8:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5361 // CHECK47-NEXT: [[TOBOOL5:%.*]] = trunc i8 [[TMP8]] to i1
5362 // CHECK47-NEXT: br i1 [[TOBOOL5]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5363 // CHECK47: omp_if.then:
5364 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5365 // CHECK47: omp.inner.for.cond:
5366 // CHECK47-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5367 // CHECK47-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
5368 // CHECK47-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
5369 // CHECK47-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5370 // CHECK47: omp.inner.for.body:
5371 // CHECK47-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5372 // CHECK47-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
5373 // CHECK47-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5374 // CHECK47-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !3
5375 // CHECK47-NEXT: [[TMP12:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !3
5376 // CHECK47-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP12]]
5377 // CHECK47-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
5378 // CHECK47-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5379 // CHECK47: omp.body.continue:
5380 // CHECK47-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5381 // CHECK47: omp.inner.for.inc:
5382 // CHECK47-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5383 // CHECK47-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP13]], 1
5384 // CHECK47-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5385 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
5386 // CHECK47: omp.inner.for.end:
5387 // CHECK47-NEXT: br label [[OMP_IF_END:%.*]]
5388 // CHECK47: omp_if.else:
5389 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
5390 // CHECK47: omp.inner.for.cond8:
5391 // CHECK47-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5392 // CHECK47-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5393 // CHECK47-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5394 // CHECK47-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
5395 // CHECK47: omp.inner.for.body10:
5396 // CHECK47-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5397 // CHECK47-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP16]], 1
5398 // CHECK47-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
5399 // CHECK47-NEXT: store i32 [[ADD12]], i32* [[I4]], align 4
5400 // CHECK47-NEXT: [[TMP17:%.*]] = load i32, i32* [[I4]], align 4
5401 // CHECK47-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP17]]
5402 // CHECK47-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4
5403 // CHECK47-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
5404 // CHECK47: omp.body.continue14:
5405 // CHECK47-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
5406 // CHECK47: omp.inner.for.inc15:
5407 // CHECK47-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5408 // CHECK47-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP18]], 1
5409 // CHECK47-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV]], align 4
5410 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
5411 // CHECK47: omp.inner.for.end17:
5412 // CHECK47-NEXT: br label [[OMP_IF_END]]
5413 // CHECK47: omp_if.end:
5414 // CHECK47-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5415 // CHECK47-NEXT: [[SUB18:%.*]] = sub nsw i32 [[TMP19]], 0
5416 // CHECK47-NEXT: [[DIV19:%.*]] = sdiv i32 [[SUB18]], 1
5417 // CHECK47-NEXT: [[MUL20:%.*]] = mul nsw i32 [[DIV19]], 1
5418 // CHECK47-NEXT: [[ADD21:%.*]] = add nsw i32 0, [[MUL20]]
5419 // CHECK47-NEXT: store i32 [[ADD21]], i32* [[I4]], align 4
5420 // CHECK47-NEXT: br label [[SIMD_IF_END]]
5421 // CHECK47: simd.if.end:
5422 // CHECK47-NEXT: [[TMP20:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5423 // CHECK47-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP20]])
5424 // CHECK47-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
5425 // CHECK47-NEXT: [[TMP21:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5426 // CHECK47-NEXT: call void @llvm.stackrestore(i8* [[TMP21]])
5427 // CHECK47-NEXT: [[TMP22:%.*]] = load i32, i32* [[RETVAL]], align 4
5428 // CHECK47-NEXT: ret i32 [[TMP22]]
5429 //
5430 //
5431 // CHECK47-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5432 // CHECK47-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
5433 // CHECK47-NEXT: entry:
5434 // CHECK47-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5435 // CHECK47-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5436 // CHECK47-NEXT: [[TE:%.*]] = alloca i32, align 4
5437 // CHECK47-NEXT: [[TH:%.*]] = alloca i32, align 4
5438 // CHECK47-NEXT: [[TMP:%.*]] = alloca i32, align 4
5439 // CHECK47-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5440 // CHECK47-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5441 // CHECK47-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5442 // CHECK47-NEXT: [[I:%.*]] = alloca i32, align 4
5443 // CHECK47-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5444 // CHECK47-NEXT: store i32 0, i32* [[TE]], align 4
5445 // CHECK47-NEXT: store i32 128, i32* [[TH]], align 4
5446 // CHECK47-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5447 // CHECK47-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5448 // CHECK47-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5449 // CHECK47-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
5450 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5451 // CHECK47: omp.inner.for.cond:
5452 // CHECK47-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
5453 // CHECK47-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
5454 // CHECK47-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
5455 // CHECK47-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5456 // CHECK47: omp.inner.for.body:
5457 // CHECK47-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
5458 // CHECK47-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
5459 // CHECK47-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5460 // CHECK47-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
5461 // CHECK47-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !9
5462 // CHECK47-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
5463 // CHECK47-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !9
5464 // CHECK47-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5465 // CHECK47: omp.body.continue:
5466 // CHECK47-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5467 // CHECK47: omp.inner.for.inc:
5468 // CHECK47-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
5469 // CHECK47-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5470 // CHECK47-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
5471 // CHECK47-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
5472 // CHECK47: omp.inner.for.end:
5473 // CHECK47-NEXT: store i32 10, i32* [[I]], align 4
5474 // CHECK47-NEXT: ret i32 0
5475 //
5476