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
6 // Test host codegen.
7 // 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
8 // 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
9 // 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
10 // 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
11 // 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
12 // 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
13
14 // 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
15 // 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
16 // 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
17 // 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
18 // 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
19 // 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
20 #ifdef CK1
21
22 template <typename T, int X, long long Y>
23 struct SS{
24 T a[X];
25 float b;
fooSS26 int foo(void) {
27
28 #pragma omp target
29 #pragma omp teams distribute parallel for simd
30 for(int i = 0; i < X; i++) {
31 a[i] = (T)0;
32 }
33 #pragma omp target
34 #pragma omp teams distribute parallel for simd dist_schedule(static)
35 for(int i = 0; i < X; i++) {
36 a[i] = (T)0;
37 }
38 #pragma omp target
39 #pragma omp teams distribute parallel for simd dist_schedule(static, X/2)
40 for(int i = 0; i < X; i++) {
41 a[i] = (T)0;
42 }
43
44
45
46
47
48
49
50
51
52
53 return a[0];
54 }
55 };
56
teams_template_struct(void)57 int teams_template_struct(void) {
58 SS<int, 123, 456> V;
59 return V.foo();
60
61 }
62
63
64 #endif // CK1
65
66 // Test host codegen.
67 // 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
68 // 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
69 // 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
70 // 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
71 // 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
72 // 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
73
74 // 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
75 // 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
76 // 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
77 // 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
78 // 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
79 // 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
80 #ifdef CK2
81
82 template <typename T, int n>
tmain(T argc)83 int tmain(T argc) {
84 T a[n];
85 int m = 10;
86 #pragma omp target
87 #pragma omp teams distribute parallel for simd
88 for(int i = 0; i < n; i++) {
89 a[i] = (T)0;
90 }
91 #pragma omp target
92 #pragma omp teams distribute parallel for simd dist_schedule(static)
93 for(int i = 0; i < n; i++) {
94 a[i] = (T)0;
95 }
96 #pragma omp target
97 #pragma omp teams distribute parallel for simd dist_schedule(static, m)
98 for(int i = 0; i < n; i++) {
99 a[i] = (T)0;
100 }
101 return 0;
102 }
103
main(int argc,char ** argv)104 int main (int argc, char **argv) {
105 int n = 100;
106 int a[n];
107 int m = 10;
108 #pragma omp target
109 #pragma omp teams distribute parallel for simd
110 for(int i = 0; i < n; i++) {
111 a[i] = 0;
112 }
113 #pragma omp target
114 #pragma omp teams distribute parallel for simd dist_schedule(static)
115 for(int i = 0; i < n; i++) {
116 a[i] = 0;
117 }
118 #pragma omp target
119 #pragma omp teams distribute parallel for simd dist_schedule(static, m)
120 for(int i = 0; i < n; i++) {
121 a[i] = 0;
122 }
123 return tmain<int, 10>(argc);
124 }
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147 #endif // CK2
148 #endif // #ifndef HEADER
149 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
150 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
151 // CHECK1-NEXT: entry:
152 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
153 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
154 // CHECK1-NEXT: ret i32 [[CALL]]
155 //
156 //
157 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
158 // CHECK1-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
159 // CHECK1-NEXT: entry:
160 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
161 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
162 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
163 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
164 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
166 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
167 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
168 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
169 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
170 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
171 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
172 // CHECK1-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
173 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
174 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
175 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
176 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
177 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
178 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
179 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
180 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
181 // CHECK1-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
182 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
183 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
184 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
185 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
186 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
187 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
188 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
189 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
190 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
191 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
192 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
193 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
194 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
195 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
196 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
197 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
198 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
199 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
200 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
201 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
202 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
203 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
204 // CHECK1-NEXT: store i64 123, i64* [[TMP15]], align 8
205 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
206 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
207 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
208 // CHECK1: omp_offload.failed:
209 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
210 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
211 // CHECK1: omp_offload.cont:
212 // CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
213 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
214 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
215 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
216 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
217 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
218 // CHECK1-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
219 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
220 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
221 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
222 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
223 // CHECK1-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
224 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
225 // CHECK1-NEXT: store i32 1, i32* [[TMP25]], align 4
226 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
227 // CHECK1-NEXT: store i32 1, i32* [[TMP26]], align 4
228 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
229 // CHECK1-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
230 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
231 // CHECK1-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
232 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
233 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 8
234 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
235 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 8
236 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
237 // CHECK1-NEXT: store i8** null, i8*** [[TMP31]], align 8
238 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
239 // CHECK1-NEXT: store i8** null, i8*** [[TMP32]], align 8
240 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
241 // CHECK1-NEXT: store i64 123, i64* [[TMP33]], align 8
242 // CHECK1-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
243 // CHECK1-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
244 // CHECK1-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
245 // CHECK1: omp_offload.failed8:
246 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
247 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT9]]
248 // CHECK1: omp_offload.cont9:
249 // CHECK1-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
250 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
251 // CHECK1-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
252 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
253 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
254 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
255 // CHECK1-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
256 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
257 // CHECK1-NEXT: store i8* null, i8** [[TMP40]], align 8
258 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
259 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
260 // CHECK1-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
261 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
262 // CHECK1-NEXT: store i32 1, i32* [[TMP43]], align 4
263 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
264 // CHECK1-NEXT: store i32 1, i32* [[TMP44]], align 4
265 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
266 // CHECK1-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 8
267 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
268 // CHECK1-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 8
269 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
270 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 8
271 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
272 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 8
273 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
274 // CHECK1-NEXT: store i8** null, i8*** [[TMP49]], align 8
275 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
276 // CHECK1-NEXT: store i8** null, i8*** [[TMP50]], align 8
277 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
278 // CHECK1-NEXT: store i64 123, i64* [[TMP51]], align 8
279 // CHECK1-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
280 // CHECK1-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
281 // CHECK1-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
282 // CHECK1: omp_offload.failed16:
283 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
284 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT17]]
285 // CHECK1: omp_offload.cont17:
286 // CHECK1-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
287 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i64 0, i64 0
288 // CHECK1-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
289 // CHECK1-NEXT: ret i32 [[TMP54]]
290 //
291 //
292 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
293 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
294 // CHECK1-NEXT: entry:
295 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
296 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
297 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
298 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
299 // CHECK1-NEXT: ret void
300 //
301 //
302 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
303 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
304 // CHECK1-NEXT: entry:
305 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
306 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
307 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
308 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
309 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
310 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
311 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
312 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
313 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
314 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
315 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
316 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
317 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
318 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
319 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
320 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
321 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
322 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
323 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
324 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
325 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
326 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
327 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
328 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
329 // CHECK1: cond.true:
330 // CHECK1-NEXT: br label [[COND_END:%.*]]
331 // CHECK1: cond.false:
332 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
333 // CHECK1-NEXT: br label [[COND_END]]
334 // CHECK1: cond.end:
335 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
336 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
337 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
338 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
339 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
340 // CHECK1: omp.inner.for.cond:
341 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
342 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !6
343 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
344 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
345 // CHECK1: omp.inner.for.body:
346 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !6
347 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
348 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !6
349 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
350 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !6
351 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
352 // CHECK1: omp.inner.for.inc:
353 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
354 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !6
355 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
356 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
357 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
358 // CHECK1: omp.inner.for.end:
359 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
360 // CHECK1: omp.loop.exit:
361 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
362 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
363 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
364 // CHECK1-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
365 // CHECK1: .omp.final.then:
366 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
367 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
368 // CHECK1: .omp.final.done:
369 // CHECK1-NEXT: ret void
370 //
371 //
372 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
373 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
374 // CHECK1-NEXT: entry:
375 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
376 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
377 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
378 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
379 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
380 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
381 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
382 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
383 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
384 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
387 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
388 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
389 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
390 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
391 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
392 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
393 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
394 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
395 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
396 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
397 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
398 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
399 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
400 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
401 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
402 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
403 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
404 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
405 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
406 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
407 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
408 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
409 // CHECK1: cond.true:
410 // CHECK1-NEXT: br label [[COND_END:%.*]]
411 // CHECK1: cond.false:
412 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
413 // CHECK1-NEXT: br label [[COND_END]]
414 // CHECK1: cond.end:
415 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
416 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
417 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
418 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
419 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
420 // CHECK1: omp.inner.for.cond:
421 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
422 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
423 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
424 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
425 // CHECK1: omp.inner.for.body:
426 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
427 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
428 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
429 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
430 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
431 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !10
432 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
433 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
434 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
435 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
436 // CHECK1: omp.body.continue:
437 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
438 // CHECK1: omp.inner.for.inc:
439 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
440 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
441 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
442 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
443 // CHECK1: omp.inner.for.end:
444 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
445 // CHECK1: omp.loop.exit:
446 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
447 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
448 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
449 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
450 // CHECK1: .omp.final.then:
451 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
452 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
453 // CHECK1: .omp.final.done:
454 // CHECK1-NEXT: ret void
455 //
456 //
457 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
458 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
459 // CHECK1-NEXT: entry:
460 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
461 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
462 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
463 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
464 // CHECK1-NEXT: ret void
465 //
466 //
467 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
468 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
469 // CHECK1-NEXT: entry:
470 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
471 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
472 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
473 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
474 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
475 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
476 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
477 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
478 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
479 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
480 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
481 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
482 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
483 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
484 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
485 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
486 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
487 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
488 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
489 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
490 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
491 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
492 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
493 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
494 // CHECK1: cond.true:
495 // CHECK1-NEXT: br label [[COND_END:%.*]]
496 // CHECK1: cond.false:
497 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
498 // CHECK1-NEXT: br label [[COND_END]]
499 // CHECK1: cond.end:
500 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
501 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
502 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
503 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
504 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
505 // CHECK1: omp.inner.for.cond:
506 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
507 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
508 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
509 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
510 // CHECK1: omp.inner.for.body:
511 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
512 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
513 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
514 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
515 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]]), !llvm.access.group !15
516 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
517 // CHECK1: omp.inner.for.inc:
518 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
519 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
520 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
521 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
522 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
523 // CHECK1: omp.inner.for.end:
524 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
525 // CHECK1: omp.loop.exit:
526 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
527 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
528 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
529 // CHECK1-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
530 // CHECK1: .omp.final.then:
531 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
532 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
533 // CHECK1: .omp.final.done:
534 // CHECK1-NEXT: ret void
535 //
536 //
537 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
538 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
539 // CHECK1-NEXT: entry:
540 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
541 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
542 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
543 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
544 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
545 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
546 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
547 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
548 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
549 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
550 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
551 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
552 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
553 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
554 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
555 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
556 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
557 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
558 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
559 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
560 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
561 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
562 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
563 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
564 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
565 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
566 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
567 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
568 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
569 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
570 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
571 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
572 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
573 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
574 // CHECK1: cond.true:
575 // CHECK1-NEXT: br label [[COND_END:%.*]]
576 // CHECK1: cond.false:
577 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
578 // CHECK1-NEXT: br label [[COND_END]]
579 // CHECK1: cond.end:
580 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
581 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
582 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
583 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
584 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
585 // CHECK1: omp.inner.for.cond:
586 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
587 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
588 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
589 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
590 // CHECK1: omp.inner.for.body:
591 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
592 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
593 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
594 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
595 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
596 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !18
597 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
598 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
599 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
600 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
601 // CHECK1: omp.body.continue:
602 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
603 // CHECK1: omp.inner.for.inc:
604 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
605 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
606 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
607 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
608 // CHECK1: omp.inner.for.end:
609 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
610 // CHECK1: omp.loop.exit:
611 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
612 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
613 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
614 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
615 // CHECK1: .omp.final.then:
616 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
617 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
618 // CHECK1: .omp.final.done:
619 // CHECK1-NEXT: ret void
620 //
621 //
622 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
623 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
624 // CHECK1-NEXT: entry:
625 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
626 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
627 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
628 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
629 // CHECK1-NEXT: ret void
630 //
631 //
632 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
633 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
634 // CHECK1-NEXT: entry:
635 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
636 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
637 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
638 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
639 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
640 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
641 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
642 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
643 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
644 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
645 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
646 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
647 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
648 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
649 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
650 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
651 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
652 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
653 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
654 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
655 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
656 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
657 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
658 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
659 // CHECK1: cond.true:
660 // CHECK1-NEXT: br label [[COND_END:%.*]]
661 // CHECK1: cond.false:
662 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
663 // CHECK1-NEXT: br label [[COND_END]]
664 // CHECK1: cond.end:
665 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
666 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
667 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
668 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
669 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
670 // CHECK1: omp.inner.for.cond:
671 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
672 // CHECK1-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
673 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
674 // CHECK1: omp.inner.for.body:
675 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !21
676 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
677 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
678 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
679 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]], %struct.SS* [[TMP0]]), !llvm.access.group !21
680 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
681 // CHECK1: omp.inner.for.inc:
682 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
683 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !21
684 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
685 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
686 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !21
687 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !21
688 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
689 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !21
690 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
691 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !21
692 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
693 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
694 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
695 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP17]], 122
696 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
697 // CHECK1: cond.true5:
698 // CHECK1-NEXT: br label [[COND_END7:%.*]]
699 // CHECK1: cond.false6:
700 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
701 // CHECK1-NEXT: br label [[COND_END7]]
702 // CHECK1: cond.end7:
703 // CHECK1-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP18]], [[COND_FALSE6]] ]
704 // CHECK1-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !21
705 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !21
706 // CHECK1-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
707 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
708 // CHECK1: omp.inner.for.end:
709 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
710 // CHECK1: omp.loop.exit:
711 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
712 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
713 // CHECK1-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
714 // CHECK1-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
715 // CHECK1: .omp.final.then:
716 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
717 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
718 // CHECK1: .omp.final.done:
719 // CHECK1-NEXT: ret void
720 //
721 //
722 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
723 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
724 // CHECK1-NEXT: entry:
725 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
726 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
727 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
728 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
729 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
730 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
731 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
732 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
733 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
734 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
735 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
736 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
737 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
738 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
739 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
740 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
741 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
742 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
743 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
744 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
745 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
746 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
747 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
748 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
749 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
750 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
751 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
752 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
753 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
754 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
755 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
756 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
757 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
758 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
759 // CHECK1: cond.true:
760 // CHECK1-NEXT: br label [[COND_END:%.*]]
761 // CHECK1: cond.false:
762 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
763 // CHECK1-NEXT: br label [[COND_END]]
764 // CHECK1: cond.end:
765 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
766 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
767 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
768 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
769 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
770 // CHECK1: omp.inner.for.cond:
771 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
772 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
773 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
774 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
775 // CHECK1: omp.inner.for.body:
776 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
777 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
778 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
779 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !24
780 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
781 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
782 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
783 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
784 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
785 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
786 // CHECK1: omp.body.continue:
787 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
788 // CHECK1: omp.inner.for.inc:
789 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
790 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
791 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
792 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
793 // CHECK1: omp.inner.for.end:
794 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
795 // CHECK1: omp.loop.exit:
796 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
797 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
798 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
799 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
800 // CHECK1: .omp.final.then:
801 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
802 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
803 // CHECK1: .omp.final.done:
804 // CHECK1-NEXT: ret void
805 //
806 //
807 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
808 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
809 // CHECK1-NEXT: entry:
810 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
811 // CHECK1-NEXT: ret void
812 //
813 //
814 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
815 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
816 // CHECK3-NEXT: entry:
817 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
818 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
819 // CHECK3-NEXT: ret i32 [[CALL]]
820 //
821 //
822 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
823 // CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
824 // CHECK3-NEXT: entry:
825 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
826 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
827 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
828 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
829 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
830 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
831 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
832 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
833 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
834 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
835 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
836 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
837 // CHECK3-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
838 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
839 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
840 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
841 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
842 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
843 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
844 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
845 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
846 // CHECK3-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
847 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
848 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
849 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
850 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
851 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
852 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
853 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
854 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
855 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
856 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
857 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
858 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
859 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
860 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
861 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
862 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
863 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
864 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
865 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
866 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
867 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
868 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
869 // CHECK3-NEXT: store i64 123, i64* [[TMP15]], align 8
870 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
871 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
872 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
873 // CHECK3: omp_offload.failed:
874 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
875 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
876 // CHECK3: omp_offload.cont:
877 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
878 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
879 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
880 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
881 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
882 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
883 // CHECK3-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
884 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
885 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
886 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
887 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
888 // CHECK3-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
889 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
890 // CHECK3-NEXT: store i32 1, i32* [[TMP25]], align 4
891 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
892 // CHECK3-NEXT: store i32 1, i32* [[TMP26]], align 4
893 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
894 // CHECK3-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
895 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
896 // CHECK3-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
897 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
898 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 4
899 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
900 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 4
901 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
902 // CHECK3-NEXT: store i8** null, i8*** [[TMP31]], align 4
903 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
904 // CHECK3-NEXT: store i8** null, i8*** [[TMP32]], align 4
905 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
906 // CHECK3-NEXT: store i64 123, i64* [[TMP33]], align 8
907 // CHECK3-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
908 // CHECK3-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
909 // CHECK3-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
910 // CHECK3: omp_offload.failed8:
911 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
912 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT9]]
913 // CHECK3: omp_offload.cont9:
914 // CHECK3-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
915 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
916 // CHECK3-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
917 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
918 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
919 // CHECK3-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
920 // CHECK3-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
921 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
922 // CHECK3-NEXT: store i8* null, i8** [[TMP40]], align 4
923 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
924 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
925 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
926 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
927 // CHECK3-NEXT: store i32 1, i32* [[TMP43]], align 4
928 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
929 // CHECK3-NEXT: store i32 1, i32* [[TMP44]], align 4
930 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
931 // CHECK3-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 4
932 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
933 // CHECK3-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 4
934 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
935 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 4
936 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
937 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 4
938 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
939 // CHECK3-NEXT: store i8** null, i8*** [[TMP49]], align 4
940 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
941 // CHECK3-NEXT: store i8** null, i8*** [[TMP50]], align 4
942 // CHECK3-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
943 // CHECK3-NEXT: store i64 123, i64* [[TMP51]], align 8
944 // CHECK3-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
945 // CHECK3-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
946 // CHECK3-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
947 // CHECK3: omp_offload.failed16:
948 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
949 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]]
950 // CHECK3: omp_offload.cont17:
951 // CHECK3-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
952 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i32 0, i32 0
953 // CHECK3-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
954 // CHECK3-NEXT: ret i32 [[TMP54]]
955 //
956 //
957 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
958 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
959 // CHECK3-NEXT: entry:
960 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
961 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
962 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
963 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
964 // CHECK3-NEXT: ret void
965 //
966 //
967 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
968 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
969 // CHECK3-NEXT: entry:
970 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
971 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
972 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
973 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
974 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
975 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
976 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
977 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
978 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
979 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
980 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
981 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
982 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
983 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
984 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
985 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
986 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
987 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
988 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
989 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
990 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
991 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
992 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
993 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
994 // CHECK3: cond.true:
995 // CHECK3-NEXT: br label [[COND_END:%.*]]
996 // CHECK3: cond.false:
997 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
998 // CHECK3-NEXT: br label [[COND_END]]
999 // CHECK3: cond.end:
1000 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1001 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1002 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1003 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1004 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1005 // CHECK3: omp.inner.for.cond:
1006 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1007 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !7
1008 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1009 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1010 // CHECK3: omp.inner.for.body:
1011 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !7
1012 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !7
1013 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !7
1014 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1015 // CHECK3: omp.inner.for.inc:
1016 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1017 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !7
1018 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1019 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1020 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
1021 // CHECK3: omp.inner.for.end:
1022 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1023 // CHECK3: omp.loop.exit:
1024 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1025 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1026 // CHECK3-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1027 // CHECK3-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1028 // CHECK3: .omp.final.then:
1029 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1030 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1031 // CHECK3: .omp.final.done:
1032 // CHECK3-NEXT: ret void
1033 //
1034 //
1035 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1036 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1037 // CHECK3-NEXT: entry:
1038 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1039 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1040 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1041 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1042 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1043 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1044 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1045 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1046 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1047 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1048 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1049 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1050 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1051 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1052 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1053 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1054 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1055 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1056 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1057 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1058 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1059 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1060 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1061 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1062 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1063 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1064 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1065 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1066 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1067 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1068 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1069 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1070 // CHECK3: cond.true:
1071 // CHECK3-NEXT: br label [[COND_END:%.*]]
1072 // CHECK3: cond.false:
1073 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1074 // CHECK3-NEXT: br label [[COND_END]]
1075 // CHECK3: cond.end:
1076 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1077 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1078 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1079 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1080 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1081 // CHECK3: omp.inner.for.cond:
1082 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1083 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
1084 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1085 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1086 // CHECK3: omp.inner.for.body:
1087 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1088 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1089 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1090 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
1091 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1092 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
1093 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1094 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
1095 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1096 // CHECK3: omp.body.continue:
1097 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1098 // CHECK3: omp.inner.for.inc:
1099 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1100 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1101 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1102 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1103 // CHECK3: omp.inner.for.end:
1104 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1105 // CHECK3: omp.loop.exit:
1106 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1107 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1108 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1109 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1110 // CHECK3: .omp.final.then:
1111 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1112 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1113 // CHECK3: .omp.final.done:
1114 // CHECK3-NEXT: ret void
1115 //
1116 //
1117 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
1118 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1119 // CHECK3-NEXT: entry:
1120 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1121 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1122 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1123 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1124 // CHECK3-NEXT: ret void
1125 //
1126 //
1127 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1128 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1129 // CHECK3-NEXT: entry:
1130 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1131 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1132 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1133 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1134 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1135 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1136 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1137 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1138 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1139 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1140 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1141 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1142 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1143 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1144 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1145 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1146 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1147 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1148 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1149 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1150 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1151 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1152 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1153 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1154 // CHECK3: cond.true:
1155 // CHECK3-NEXT: br label [[COND_END:%.*]]
1156 // CHECK3: cond.false:
1157 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1158 // CHECK3-NEXT: br label [[COND_END]]
1159 // CHECK3: cond.end:
1160 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1161 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1162 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1163 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1164 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1165 // CHECK3: omp.inner.for.cond:
1166 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
1167 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !16
1168 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1169 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1170 // CHECK3: omp.inner.for.body:
1171 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !16
1172 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !16
1173 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]]), !llvm.access.group !16
1174 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1175 // CHECK3: omp.inner.for.inc:
1176 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
1177 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !16
1178 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1179 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
1180 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
1181 // CHECK3: omp.inner.for.end:
1182 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1183 // CHECK3: omp.loop.exit:
1184 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1185 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1186 // CHECK3-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1187 // CHECK3-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1188 // CHECK3: .omp.final.then:
1189 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1190 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1191 // CHECK3: .omp.final.done:
1192 // CHECK3-NEXT: ret void
1193 //
1194 //
1195 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1196 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1197 // CHECK3-NEXT: entry:
1198 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1199 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1200 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1201 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1202 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1203 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1204 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1205 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1206 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1207 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1208 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1209 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1210 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1211 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1212 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1213 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1214 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1215 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1216 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1217 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1218 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1219 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1220 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1221 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1222 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1223 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1224 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1225 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1226 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1227 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1228 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1229 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1230 // CHECK3: cond.true:
1231 // CHECK3-NEXT: br label [[COND_END:%.*]]
1232 // CHECK3: cond.false:
1233 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1234 // CHECK3-NEXT: br label [[COND_END]]
1235 // CHECK3: cond.end:
1236 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1237 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1238 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1239 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1240 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1241 // CHECK3: omp.inner.for.cond:
1242 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
1243 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
1244 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1245 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1246 // CHECK3: omp.inner.for.body:
1247 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
1248 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1249 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1250 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
1251 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1252 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !19
1253 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1254 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
1255 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1256 // CHECK3: omp.body.continue:
1257 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1258 // CHECK3: omp.inner.for.inc:
1259 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
1260 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1261 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
1262 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
1263 // CHECK3: omp.inner.for.end:
1264 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1265 // CHECK3: omp.loop.exit:
1266 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1267 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1268 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1269 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1270 // CHECK3: .omp.final.then:
1271 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1272 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1273 // CHECK3: .omp.final.done:
1274 // CHECK3-NEXT: ret void
1275 //
1276 //
1277 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
1278 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1279 // CHECK3-NEXT: entry:
1280 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1281 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1282 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1283 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1284 // CHECK3-NEXT: ret void
1285 //
1286 //
1287 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
1288 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1289 // CHECK3-NEXT: entry:
1290 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1291 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1292 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1293 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1294 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1295 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1296 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1297 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1298 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1299 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1300 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1301 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1302 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1303 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1304 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1305 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1306 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1307 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1308 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1309 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1310 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
1311 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1312 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1313 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1314 // CHECK3: cond.true:
1315 // CHECK3-NEXT: br label [[COND_END:%.*]]
1316 // CHECK3: cond.false:
1317 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1318 // CHECK3-NEXT: br label [[COND_END]]
1319 // CHECK3: cond.end:
1320 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1321 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1322 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1323 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1324 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1325 // CHECK3: omp.inner.for.cond:
1326 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1327 // CHECK3-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
1328 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1329 // CHECK3: omp.inner.for.body:
1330 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
1331 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1332 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP7]], i32 [[TMP8]], %struct.SS* [[TMP0]]), !llvm.access.group !22
1333 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1334 // CHECK3: omp.inner.for.inc:
1335 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1336 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
1337 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
1338 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1339 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
1340 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
1341 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1342 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
1343 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1344 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !22
1345 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1346 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1347 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1348 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP15]], 122
1349 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
1350 // CHECK3: cond.true5:
1351 // CHECK3-NEXT: br label [[COND_END7:%.*]]
1352 // CHECK3: cond.false6:
1353 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1354 // CHECK3-NEXT: br label [[COND_END7]]
1355 // CHECK3: cond.end7:
1356 // CHECK3-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP16]], [[COND_FALSE6]] ]
1357 // CHECK3-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !22
1358 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !22
1359 // CHECK3-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
1360 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
1361 // CHECK3: omp.inner.for.end:
1362 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1363 // CHECK3: omp.loop.exit:
1364 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1365 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1366 // CHECK3-NEXT: [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1367 // CHECK3-NEXT: br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1368 // CHECK3: .omp.final.then:
1369 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1370 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1371 // CHECK3: .omp.final.done:
1372 // CHECK3-NEXT: ret void
1373 //
1374 //
1375 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1376 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1377 // CHECK3-NEXT: entry:
1378 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1379 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1380 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1381 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1382 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1383 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1384 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1385 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1386 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1387 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1388 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1389 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1390 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1391 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1392 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1393 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1394 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1395 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1396 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1397 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1398 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1399 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1400 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1401 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1402 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1403 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1404 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1405 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1406 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1407 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1408 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1409 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1410 // CHECK3: cond.true:
1411 // CHECK3-NEXT: br label [[COND_END:%.*]]
1412 // CHECK3: cond.false:
1413 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1414 // CHECK3-NEXT: br label [[COND_END]]
1415 // CHECK3: cond.end:
1416 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1417 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1418 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1419 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1420 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1421 // CHECK3: omp.inner.for.cond:
1422 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
1423 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
1424 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1425 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1426 // CHECK3: omp.inner.for.body:
1427 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
1428 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1429 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1430 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !25
1431 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1432 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
1433 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1434 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
1435 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1436 // CHECK3: omp.body.continue:
1437 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1438 // CHECK3: omp.inner.for.inc:
1439 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
1440 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1441 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
1442 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
1443 // CHECK3: omp.inner.for.end:
1444 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1445 // CHECK3: omp.loop.exit:
1446 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1447 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1448 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1449 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1450 // CHECK3: .omp.final.then:
1451 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1452 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1453 // CHECK3: .omp.final.done:
1454 // CHECK3-NEXT: ret void
1455 //
1456 //
1457 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1458 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1459 // CHECK3-NEXT: entry:
1460 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1461 // CHECK3-NEXT: ret void
1462 //
1463 //
1464 // CHECK5-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1465 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
1466 // CHECK5-NEXT: entry:
1467 // CHECK5-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1468 // CHECK5-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
1469 // CHECK5-NEXT: ret i32 [[CALL]]
1470 //
1471 //
1472 // CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1473 // CHECK5-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1474 // CHECK5-NEXT: entry:
1475 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1476 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
1477 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1478 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1479 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1480 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
1481 // CHECK5-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
1482 // CHECK5-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
1483 // CHECK5-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
1484 // CHECK5-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
1485 // CHECK5-NEXT: [[I7:%.*]] = alloca i32, align 4
1486 // CHECK5-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
1487 // CHECK5-NEXT: [[DOTOMP_LB21:%.*]] = alloca i32, align 4
1488 // CHECK5-NEXT: [[DOTOMP_UB22:%.*]] = alloca i32, align 4
1489 // CHECK5-NEXT: [[DOTOMP_IV23:%.*]] = alloca i32, align 4
1490 // CHECK5-NEXT: [[I24:%.*]] = alloca i32, align 4
1491 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1492 // CHECK5-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1493 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1494 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1495 // CHECK5-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1496 // CHECK5-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1497 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1498 // CHECK5: omp.inner.for.cond:
1499 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1500 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1501 // CHECK5-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1502 // CHECK5-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1503 // CHECK5: omp.inner.for.body:
1504 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1505 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1506 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1507 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1508 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1509 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
1510 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
1511 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1512 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
1513 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1514 // CHECK5: omp.body.continue:
1515 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1516 // CHECK5: omp.inner.for.inc:
1517 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1518 // CHECK5-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
1519 // CHECK5-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1520 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1521 // CHECK5: omp.inner.for.end:
1522 // CHECK5-NEXT: store i32 123, i32* [[I]], align 4
1523 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
1524 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
1525 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
1526 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
1527 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
1528 // CHECK5: omp.inner.for.cond8:
1529 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
1530 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
1531 // CHECK5-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1532 // CHECK5-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
1533 // CHECK5: omp.inner.for.body10:
1534 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
1535 // CHECK5-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
1536 // CHECK5-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
1537 // CHECK5-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
1538 // CHECK5-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1539 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
1540 // CHECK5-NEXT: [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
1541 // CHECK5-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
1542 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
1543 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE16:%.*]]
1544 // CHECK5: omp.body.continue16:
1545 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC17:%.*]]
1546 // CHECK5: omp.inner.for.inc17:
1547 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
1548 // CHECK5-NEXT: [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
1549 // CHECK5-NEXT: store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
1550 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
1551 // CHECK5: omp.inner.for.end19:
1552 // CHECK5-NEXT: store i32 123, i32* [[I7]], align 4
1553 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB21]], align 4
1554 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB22]], align 4
1555 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
1556 // CHECK5-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
1557 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND25:%.*]]
1558 // CHECK5: omp.inner.for.cond25:
1559 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
1560 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
1561 // CHECK5-NEXT: [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1562 // CHECK5-NEXT: br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
1563 // CHECK5: omp.inner.for.body27:
1564 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
1565 // CHECK5-NEXT: [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
1566 // CHECK5-NEXT: [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
1567 // CHECK5-NEXT: store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
1568 // CHECK5-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1569 // CHECK5-NEXT: [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
1570 // CHECK5-NEXT: [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
1571 // CHECK5-NEXT: [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
1572 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
1573 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE33:%.*]]
1574 // CHECK5: omp.body.continue33:
1575 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC34:%.*]]
1576 // CHECK5: omp.inner.for.inc34:
1577 // CHECK5-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
1578 // CHECK5-NEXT: [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
1579 // CHECK5-NEXT: store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
1580 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
1581 // CHECK5: omp.inner.for.end36:
1582 // CHECK5-NEXT: store i32 123, i32* [[I24]], align 4
1583 // CHECK5-NEXT: [[A37:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1584 // CHECK5-NEXT: [[ARRAYIDX38:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A37]], i64 0, i64 0
1585 // CHECK5-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX38]], align 4
1586 // CHECK5-NEXT: ret i32 [[TMP18]]
1587 //
1588 //
1589 // CHECK7-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1590 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
1591 // CHECK7-NEXT: entry:
1592 // CHECK7-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1593 // CHECK7-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
1594 // CHECK7-NEXT: ret i32 [[CALL]]
1595 //
1596 //
1597 // CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1598 // CHECK7-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1599 // CHECK7-NEXT: entry:
1600 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1601 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
1602 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1603 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1604 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1605 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
1606 // CHECK7-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
1607 // CHECK7-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
1608 // CHECK7-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
1609 // CHECK7-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
1610 // CHECK7-NEXT: [[I7:%.*]] = alloca i32, align 4
1611 // CHECK7-NEXT: [[_TMP19:%.*]] = alloca i32, align 4
1612 // CHECK7-NEXT: [[DOTOMP_LB20:%.*]] = alloca i32, align 4
1613 // CHECK7-NEXT: [[DOTOMP_UB21:%.*]] = alloca i32, align 4
1614 // CHECK7-NEXT: [[DOTOMP_IV22:%.*]] = alloca i32, align 4
1615 // CHECK7-NEXT: [[I23:%.*]] = alloca i32, align 4
1616 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1617 // CHECK7-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1618 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1619 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1620 // CHECK7-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1621 // CHECK7-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1622 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1623 // CHECK7: omp.inner.for.cond:
1624 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1625 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
1626 // CHECK7-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1627 // CHECK7-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1628 // CHECK7: omp.inner.for.body:
1629 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1630 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1631 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1632 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
1633 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1634 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
1635 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
1636 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
1637 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1638 // CHECK7: omp.body.continue:
1639 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1640 // CHECK7: omp.inner.for.inc:
1641 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1642 // CHECK7-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
1643 // CHECK7-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
1644 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
1645 // CHECK7: omp.inner.for.end:
1646 // CHECK7-NEXT: store i32 123, i32* [[I]], align 4
1647 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
1648 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
1649 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
1650 // CHECK7-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
1651 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
1652 // CHECK7: omp.inner.for.cond8:
1653 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
1654 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
1655 // CHECK7-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1656 // CHECK7-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
1657 // CHECK7: omp.inner.for.body10:
1658 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
1659 // CHECK7-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
1660 // CHECK7-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
1661 // CHECK7-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
1662 // CHECK7-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1663 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
1664 // CHECK7-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
1665 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
1666 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
1667 // CHECK7: omp.body.continue15:
1668 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
1669 // CHECK7: omp.inner.for.inc16:
1670 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
1671 // CHECK7-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
1672 // CHECK7-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
1673 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
1674 // CHECK7: omp.inner.for.end18:
1675 // CHECK7-NEXT: store i32 123, i32* [[I7]], align 4
1676 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB20]], align 4
1677 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB21]], align 4
1678 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
1679 // CHECK7-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
1680 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND24:%.*]]
1681 // CHECK7: omp.inner.for.cond24:
1682 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
1683 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
1684 // CHECK7-NEXT: [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1685 // CHECK7-NEXT: br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
1686 // CHECK7: omp.inner.for.body26:
1687 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
1688 // CHECK7-NEXT: [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
1689 // CHECK7-NEXT: [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
1690 // CHECK7-NEXT: store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
1691 // CHECK7-NEXT: [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1692 // CHECK7-NEXT: [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
1693 // CHECK7-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
1694 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
1695 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE31:%.*]]
1696 // CHECK7: omp.body.continue31:
1697 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC32:%.*]]
1698 // CHECK7: omp.inner.for.inc32:
1699 // CHECK7-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
1700 // CHECK7-NEXT: [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
1701 // CHECK7-NEXT: store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
1702 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
1703 // CHECK7: omp.inner.for.end34:
1704 // CHECK7-NEXT: store i32 123, i32* [[I23]], align 4
1705 // CHECK7-NEXT: [[A35:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1706 // CHECK7-NEXT: [[ARRAYIDX36:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A35]], i32 0, i32 0
1707 // CHECK7-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX36]], align 4
1708 // CHECK7-NEXT: ret i32 [[TMP18]]
1709 //
1710 //
1711 // CHECK9-LABEL: define {{[^@]+}}@main
1712 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1713 // CHECK9-NEXT: entry:
1714 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1715 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1716 // CHECK9-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1717 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
1718 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1719 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1720 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
1721 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1722 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1723 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1724 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1725 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1726 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1727 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1728 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1729 // CHECK9-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
1730 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
1731 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
1732 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
1733 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
1734 // CHECK9-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
1735 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
1736 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
1737 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
1738 // CHECK9-NEXT: [[N_CASTED20:%.*]] = alloca i64, align 8
1739 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [4 x i8*], align 8
1740 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS23:%.*]] = alloca [4 x i8*], align 8
1741 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [4 x i8*], align 8
1742 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES25:%.*]] = alloca [4 x i64], align 8
1743 // CHECK9-NEXT: [[_TMP26:%.*]] = alloca i32, align 4
1744 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
1745 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_28:%.*]] = alloca i32, align 4
1746 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
1747 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1748 // CHECK9-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1749 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
1750 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1751 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1752 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1753 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1754 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1755 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1756 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
1757 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1758 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1759 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1760 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1761 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1762 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1763 // 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)
1764 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1765 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1766 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
1767 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1768 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1769 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
1770 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1771 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
1772 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1773 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1774 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1775 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1776 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1777 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1778 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1779 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8
1780 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1781 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1782 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1783 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1784 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1785 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1786 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1787 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1788 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1789 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1790 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1791 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1792 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1793 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1794 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1795 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1796 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1797 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1798 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1799 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1800 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1801 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1802 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1803 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1804 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1805 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4
1806 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1807 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4
1808 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1809 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1810 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1811 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1812 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1813 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1814 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1815 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1816 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1817 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8
1818 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1819 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8
1820 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1821 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1822 // CHECK9-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1823 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1824 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1825 // CHECK9: omp_offload.failed:
1826 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1827 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1828 // CHECK9: omp_offload.cont:
1829 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
1830 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
1831 // CHECK9-NEXT: store i32 [[TMP41]], i32* [[CONV4]], align 4
1832 // CHECK9-NEXT: [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
1833 // CHECK9-NEXT: [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
1834 // CHECK9-NEXT: [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
1835 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP44]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i64 24, i1 false)
1836 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1837 // CHECK9-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
1838 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP46]], align 8
1839 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1840 // CHECK9-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
1841 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP48]], align 8
1842 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
1843 // CHECK9-NEXT: store i8* null, i8** [[TMP49]], align 8
1844 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
1845 // CHECK9-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
1846 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP51]], align 8
1847 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
1848 // CHECK9-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
1849 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP53]], align 8
1850 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
1851 // CHECK9-NEXT: store i8* null, i8** [[TMP54]], align 8
1852 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
1853 // CHECK9-NEXT: [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
1854 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP56]], align 8
1855 // CHECK9-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
1856 // CHECK9-NEXT: [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
1857 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP58]], align 8
1858 // CHECK9-NEXT: [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
1859 // CHECK9-NEXT: store i64 [[TMP43]], i64* [[TMP59]], align 8
1860 // CHECK9-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
1861 // CHECK9-NEXT: store i8* null, i8** [[TMP60]], align 8
1862 // CHECK9-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1863 // CHECK9-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1864 // CHECK9-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
1865 // CHECK9-NEXT: [[TMP64:%.*]] = load i32, i32* [[N]], align 4
1866 // CHECK9-NEXT: store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
1867 // CHECK9-NEXT: [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
1868 // CHECK9-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 0
1869 // CHECK9-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
1870 // CHECK9-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
1871 // CHECK9-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
1872 // CHECK9-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
1873 // CHECK9-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
1874 // CHECK9-NEXT: [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
1875 // CHECK9-NEXT: [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1876 // CHECK9-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
1877 // CHECK9-NEXT: store i32 1, i32* [[TMP68]], align 4
1878 // CHECK9-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
1879 // CHECK9-NEXT: store i32 3, i32* [[TMP69]], align 4
1880 // CHECK9-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
1881 // CHECK9-NEXT: store i8** [[TMP61]], i8*** [[TMP70]], align 8
1882 // CHECK9-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
1883 // CHECK9-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 8
1884 // CHECK9-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
1885 // CHECK9-NEXT: store i64* [[TMP63]], i64** [[TMP72]], align 8
1886 // CHECK9-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
1887 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
1888 // CHECK9-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
1889 // CHECK9-NEXT: store i8** null, i8*** [[TMP74]], align 8
1890 // CHECK9-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
1891 // CHECK9-NEXT: store i8** null, i8*** [[TMP75]], align 8
1892 // CHECK9-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
1893 // CHECK9-NEXT: store i64 [[TMP67]], i64* [[TMP76]], align 8
1894 // CHECK9-NEXT: [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
1895 // CHECK9-NEXT: [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
1896 // CHECK9-NEXT: br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
1897 // CHECK9: omp_offload.failed17:
1898 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1899 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT18]]
1900 // CHECK9: omp_offload.cont18:
1901 // CHECK9-NEXT: [[TMP79:%.*]] = load i32, i32* [[M]], align 4
1902 // CHECK9-NEXT: [[CONV19:%.*]] = bitcast i64* [[M_CASTED]] to i32*
1903 // CHECK9-NEXT: store i32 [[TMP79]], i32* [[CONV19]], align 4
1904 // CHECK9-NEXT: [[TMP80:%.*]] = load i64, i64* [[M_CASTED]], align 8
1905 // CHECK9-NEXT: [[TMP81:%.*]] = load i32, i32* [[N]], align 4
1906 // CHECK9-NEXT: [[CONV21:%.*]] = bitcast i64* [[N_CASTED20]] to i32*
1907 // CHECK9-NEXT: store i32 [[TMP81]], i32* [[CONV21]], align 4
1908 // CHECK9-NEXT: [[TMP82:%.*]] = load i64, i64* [[N_CASTED20]], align 8
1909 // CHECK9-NEXT: [[TMP83:%.*]] = mul nuw i64 [[TMP1]], 4
1910 // CHECK9-NEXT: [[TMP84:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES25]] to i8*
1911 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP84]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i64 32, i1 false)
1912 // CHECK9-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
1913 // CHECK9-NEXT: [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
1914 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP86]], align 8
1915 // CHECK9-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
1916 // CHECK9-NEXT: [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
1917 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP88]], align 8
1918 // CHECK9-NEXT: [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
1919 // CHECK9-NEXT: store i8* null, i8** [[TMP89]], align 8
1920 // CHECK9-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
1921 // CHECK9-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
1922 // CHECK9-NEXT: store i64 [[TMP82]], i64* [[TMP91]], align 8
1923 // CHECK9-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
1924 // CHECK9-NEXT: [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i64*
1925 // CHECK9-NEXT: store i64 [[TMP82]], i64* [[TMP93]], align 8
1926 // CHECK9-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
1927 // CHECK9-NEXT: store i8* null, i8** [[TMP94]], align 8
1928 // CHECK9-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 2
1929 // CHECK9-NEXT: [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
1930 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP96]], align 8
1931 // CHECK9-NEXT: [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 2
1932 // CHECK9-NEXT: [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
1933 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP98]], align 8
1934 // CHECK9-NEXT: [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 2
1935 // CHECK9-NEXT: store i8* null, i8** [[TMP99]], align 8
1936 // CHECK9-NEXT: [[TMP100:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 3
1937 // CHECK9-NEXT: [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32**
1938 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP101]], align 8
1939 // CHECK9-NEXT: [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 3
1940 // CHECK9-NEXT: [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
1941 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP103]], align 8
1942 // CHECK9-NEXT: [[TMP104:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 3
1943 // CHECK9-NEXT: store i64 [[TMP83]], i64* [[TMP104]], align 8
1944 // CHECK9-NEXT: [[TMP105:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 3
1945 // CHECK9-NEXT: store i8* null, i8** [[TMP105]], align 8
1946 // CHECK9-NEXT: [[TMP106:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
1947 // CHECK9-NEXT: [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
1948 // CHECK9-NEXT: [[TMP108:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 0
1949 // CHECK9-NEXT: [[TMP109:%.*]] = load i32, i32* [[N]], align 4
1950 // CHECK9-NEXT: store i32 [[TMP109]], i32* [[DOTCAPTURE_EXPR_27]], align 4
1951 // CHECK9-NEXT: [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
1952 // CHECK9-NEXT: [[SUB29:%.*]] = sub nsw i32 [[TMP110]], 0
1953 // CHECK9-NEXT: [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
1954 // CHECK9-NEXT: [[SUB31:%.*]] = sub nsw i32 [[DIV30]], 1
1955 // CHECK9-NEXT: store i32 [[SUB31]], i32* [[DOTCAPTURE_EXPR_28]], align 4
1956 // CHECK9-NEXT: [[TMP111:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_28]], align 4
1957 // CHECK9-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP111]], 1
1958 // CHECK9-NEXT: [[TMP112:%.*]] = zext i32 [[ADD32]] to i64
1959 // CHECK9-NEXT: [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1960 // CHECK9-NEXT: [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 0
1961 // CHECK9-NEXT: store i32 1, i32* [[TMP113]], align 4
1962 // CHECK9-NEXT: [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 1
1963 // CHECK9-NEXT: store i32 4, i32* [[TMP114]], align 4
1964 // CHECK9-NEXT: [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 2
1965 // CHECK9-NEXT: store i8** [[TMP106]], i8*** [[TMP115]], align 8
1966 // CHECK9-NEXT: [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 3
1967 // CHECK9-NEXT: store i8** [[TMP107]], i8*** [[TMP116]], align 8
1968 // CHECK9-NEXT: [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 4
1969 // CHECK9-NEXT: store i64* [[TMP108]], i64** [[TMP117]], align 8
1970 // CHECK9-NEXT: [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 5
1971 // CHECK9-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP118]], align 8
1972 // CHECK9-NEXT: [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 6
1973 // CHECK9-NEXT: store i8** null, i8*** [[TMP119]], align 8
1974 // CHECK9-NEXT: [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 7
1975 // CHECK9-NEXT: store i8** null, i8*** [[TMP120]], align 8
1976 // CHECK9-NEXT: [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 8
1977 // CHECK9-NEXT: store i64 [[TMP112]], i64* [[TMP121]], align 8
1978 // CHECK9-NEXT: [[TMP122:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]])
1979 // CHECK9-NEXT: [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
1980 // CHECK9-NEXT: br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
1981 // CHECK9: omp_offload.failed34:
1982 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i64 [[TMP80]], i64 [[TMP82]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1983 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT35]]
1984 // CHECK9: omp_offload.cont35:
1985 // CHECK9-NEXT: [[TMP124:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1986 // CHECK9-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP124]])
1987 // CHECK9-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1988 // CHECK9-NEXT: [[TMP125:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1989 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP125]])
1990 // CHECK9-NEXT: [[TMP126:%.*]] = load i32, i32* [[RETVAL]], align 4
1991 // CHECK9-NEXT: ret i32 [[TMP126]]
1992 //
1993 //
1994 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
1995 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1996 // CHECK9-NEXT: entry:
1997 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1998 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1999 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2000 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
2001 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2002 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2003 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
2004 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2005 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2006 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
2007 // CHECK9-NEXT: ret void
2008 //
2009 //
2010 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
2011 // 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]] {
2012 // CHECK9-NEXT: entry:
2013 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2014 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2015 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2016 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2017 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2018 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2019 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2020 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2021 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2022 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2023 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2024 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2025 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2026 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2027 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
2028 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2029 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2030 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2031 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2032 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2033 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2034 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2035 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2036 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2037 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2038 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2039 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2040 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2041 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2042 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2043 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2044 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2045 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2046 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2047 // CHECK9: omp.precond.then:
2048 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2049 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2050 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
2051 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2052 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2053 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2054 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2055 // 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_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2056 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2057 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2058 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2059 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2060 // CHECK9: cond.true:
2061 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2062 // CHECK9-NEXT: br label [[COND_END:%.*]]
2063 // CHECK9: cond.false:
2064 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2065 // CHECK9-NEXT: br label [[COND_END]]
2066 // CHECK9: cond.end:
2067 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2068 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2069 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2070 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2071 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2072 // CHECK9: omp.inner.for.cond:
2073 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
2074 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
2075 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2076 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2077 // CHECK9: omp.inner.for.body:
2078 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !9
2079 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
2080 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !9
2081 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
2082 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !9
2083 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2084 // CHECK9: omp.inner.for.inc:
2085 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
2086 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !9
2087 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
2088 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
2089 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
2090 // CHECK9: omp.inner.for.end:
2091 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2092 // CHECK9: omp.loop.exit:
2093 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2094 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
2095 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
2096 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2097 // CHECK9-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
2098 // CHECK9-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2099 // CHECK9: .omp.final.then:
2100 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2101 // CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
2102 // CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
2103 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
2104 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
2105 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
2106 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2107 // CHECK9: .omp.final.done:
2108 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2109 // CHECK9: omp.precond.end:
2110 // CHECK9-NEXT: ret void
2111 //
2112 //
2113 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
2114 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2115 // CHECK9-NEXT: entry:
2116 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2117 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2118 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2119 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2120 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2121 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2122 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2123 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2124 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2125 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2126 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2127 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2128 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2129 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2130 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2131 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2132 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
2133 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2134 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2135 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2136 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2137 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2138 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2139 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2140 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2141 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2142 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2143 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2144 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2145 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2146 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2147 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2148 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2149 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2150 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2151 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2152 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2153 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2154 // CHECK9: omp.precond.then:
2155 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2156 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2157 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2158 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2159 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
2160 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2161 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
2162 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2163 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
2164 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2165 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2166 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2167 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2168 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2169 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2170 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2171 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2172 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2173 // CHECK9: cond.true:
2174 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2175 // CHECK9-NEXT: br label [[COND_END:%.*]]
2176 // CHECK9: cond.false:
2177 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2178 // CHECK9-NEXT: br label [[COND_END]]
2179 // CHECK9: cond.end:
2180 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2181 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2182 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2183 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
2184 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2185 // CHECK9: omp.inner.for.cond:
2186 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2187 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
2188 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2189 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2190 // CHECK9: omp.inner.for.body:
2191 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2192 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2193 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2194 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !13
2195 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !13
2196 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2197 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
2198 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
2199 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2200 // CHECK9: omp.body.continue:
2201 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2202 // CHECK9: omp.inner.for.inc:
2203 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2204 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
2205 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2206 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
2207 // CHECK9: omp.inner.for.end:
2208 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2209 // CHECK9: omp.loop.exit:
2210 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2211 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2212 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2213 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2214 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
2215 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2216 // CHECK9: .omp.final.then:
2217 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2218 // CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
2219 // CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
2220 // CHECK9-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
2221 // CHECK9-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
2222 // CHECK9-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
2223 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2224 // CHECK9: .omp.final.done:
2225 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2226 // CHECK9: omp.precond.end:
2227 // CHECK9-NEXT: ret void
2228 //
2229 //
2230 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
2231 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2232 // CHECK9-NEXT: entry:
2233 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
2234 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2235 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2236 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
2237 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2238 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2239 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
2240 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2241 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2242 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
2243 // CHECK9-NEXT: ret void
2244 //
2245 //
2246 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
2247 // 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]] {
2248 // CHECK9-NEXT: entry:
2249 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2250 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2251 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2252 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2253 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2254 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2255 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2256 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2257 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2258 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2259 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2260 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2261 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2262 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2263 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
2264 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2265 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2266 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2267 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2268 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2269 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2270 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2271 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2272 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2273 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2274 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2275 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2276 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2277 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2278 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2279 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2280 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2281 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2282 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2283 // CHECK9: omp.precond.then:
2284 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2285 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2286 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
2287 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2288 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2289 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2290 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2291 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2292 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2293 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2294 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2295 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2296 // CHECK9: cond.true:
2297 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2298 // CHECK9-NEXT: br label [[COND_END:%.*]]
2299 // CHECK9: cond.false:
2300 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2301 // CHECK9-NEXT: br label [[COND_END]]
2302 // CHECK9: cond.end:
2303 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2304 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2305 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2306 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2307 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2308 // CHECK9: omp.inner.for.cond:
2309 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
2310 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
2311 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2312 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2313 // CHECK9: omp.inner.for.body:
2314 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !18
2315 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
2316 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !18
2317 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
2318 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !18
2319 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2320 // CHECK9: omp.inner.for.inc:
2321 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
2322 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !18
2323 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
2324 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
2325 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
2326 // CHECK9: omp.inner.for.end:
2327 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2328 // CHECK9: omp.loop.exit:
2329 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2330 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
2331 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
2332 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2333 // CHECK9-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
2334 // CHECK9-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2335 // CHECK9: .omp.final.then:
2336 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2337 // CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
2338 // CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
2339 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
2340 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
2341 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
2342 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2343 // CHECK9: .omp.final.done:
2344 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2345 // CHECK9: omp.precond.end:
2346 // CHECK9-NEXT: ret void
2347 //
2348 //
2349 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
2350 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2351 // CHECK9-NEXT: entry:
2352 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2353 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2354 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2355 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2356 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2357 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2358 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2359 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2360 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2361 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2362 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2363 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2364 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2365 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2366 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2367 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2368 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
2369 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2370 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2371 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2372 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2373 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2374 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2375 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2376 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2377 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2378 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2379 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2380 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2381 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2382 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2383 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2384 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2385 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2386 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2387 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2388 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2389 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2390 // CHECK9: omp.precond.then:
2391 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2392 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2393 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2394 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2395 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
2396 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2397 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
2398 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2399 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
2400 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2401 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2402 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2403 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2404 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2405 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2406 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2407 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2408 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2409 // CHECK9: cond.true:
2410 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2411 // CHECK9-NEXT: br label [[COND_END:%.*]]
2412 // CHECK9: cond.false:
2413 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2414 // CHECK9-NEXT: br label [[COND_END]]
2415 // CHECK9: cond.end:
2416 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2417 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2418 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2419 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
2420 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2421 // CHECK9: omp.inner.for.cond:
2422 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
2423 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
2424 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2425 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2426 // CHECK9: omp.inner.for.body:
2427 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
2428 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2429 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2430 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !21
2431 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !21
2432 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2433 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
2434 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
2435 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2436 // CHECK9: omp.body.continue:
2437 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2438 // CHECK9: omp.inner.for.inc:
2439 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
2440 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
2441 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
2442 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
2443 // CHECK9: omp.inner.for.end:
2444 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2445 // CHECK9: omp.loop.exit:
2446 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2447 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2448 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2449 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2450 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
2451 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2452 // CHECK9: .omp.final.then:
2453 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2454 // CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
2455 // CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
2456 // CHECK9-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
2457 // CHECK9-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
2458 // CHECK9-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
2459 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2460 // CHECK9: .omp.final.done:
2461 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2462 // CHECK9: omp.precond.end:
2463 // CHECK9-NEXT: ret void
2464 //
2465 //
2466 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
2467 // CHECK9-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2468 // CHECK9-NEXT: entry:
2469 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
2470 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
2471 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2472 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2473 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2474 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2475 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
2476 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
2477 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2478 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2479 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
2480 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
2481 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2482 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2483 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
2484 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
2485 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2486 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2487 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
2488 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2489 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
2490 // CHECK9-NEXT: ret void
2491 //
2492 //
2493 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
2494 // 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:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2495 // CHECK9-NEXT: entry:
2496 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2497 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2498 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2499 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2500 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2501 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2502 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2503 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2504 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2505 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2506 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2507 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2508 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2509 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2510 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2511 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
2512 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2513 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2514 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2515 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2516 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2517 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2518 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2519 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2520 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2521 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2522 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2523 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2524 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2525 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2526 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2527 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2528 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2529 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2530 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2531 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2532 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2533 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2534 // CHECK9: omp.precond.then:
2535 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2536 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2537 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
2538 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2539 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2540 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 4
2541 // CHECK9-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2542 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
2543 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
2544 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2545 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2546 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
2547 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2548 // CHECK9: cond.true:
2549 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2550 // CHECK9-NEXT: br label [[COND_END:%.*]]
2551 // CHECK9: cond.false:
2552 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2553 // CHECK9-NEXT: br label [[COND_END]]
2554 // CHECK9: cond.end:
2555 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
2556 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2557 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2558 // CHECK9-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
2559 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2560 // CHECK9: omp.inner.for.cond:
2561 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
2562 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !24
2563 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
2564 // CHECK9-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
2565 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2566 // CHECK9: omp.inner.for.body:
2567 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
2568 // CHECK9-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
2569 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2570 // CHECK9-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
2571 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !24
2572 // CHECK9-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2573 // CHECK9-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4, !llvm.access.group !24
2574 // CHECK9-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !24
2575 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]]), !llvm.access.group !24
2576 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2577 // CHECK9: omp.inner.for.inc:
2578 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
2579 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
2580 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
2581 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
2582 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
2583 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
2584 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
2585 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
2586 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2587 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
2588 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
2589 // CHECK9-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2590 // CHECK9-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2591 // CHECK9-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !24
2592 // CHECK9-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
2593 // CHECK9-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
2594 // CHECK9: cond.true12:
2595 // CHECK9-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !24
2596 // CHECK9-NEXT: br label [[COND_END14:%.*]]
2597 // CHECK9: cond.false13:
2598 // CHECK9-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2599 // CHECK9-NEXT: br label [[COND_END14]]
2600 // CHECK9: cond.end14:
2601 // CHECK9-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
2602 // CHECK9-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
2603 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
2604 // CHECK9-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
2605 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
2606 // CHECK9: omp.inner.for.end:
2607 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2608 // CHECK9: omp.loop.exit:
2609 // CHECK9-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2610 // CHECK9-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
2611 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
2612 // CHECK9-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2613 // CHECK9-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
2614 // CHECK9-NEXT: br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2615 // CHECK9: .omp.final.then:
2616 // CHECK9-NEXT: [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2617 // CHECK9-NEXT: [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
2618 // CHECK9-NEXT: [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
2619 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
2620 // CHECK9-NEXT: [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
2621 // CHECK9-NEXT: store i32 [[ADD18]], i32* [[I4]], align 4
2622 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2623 // CHECK9: .omp.final.done:
2624 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2625 // CHECK9: omp.precond.end:
2626 // CHECK9-NEXT: ret void
2627 //
2628 //
2629 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
2630 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2631 // CHECK9-NEXT: entry:
2632 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2633 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2634 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2635 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2636 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2637 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2638 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2639 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2640 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2641 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2642 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2643 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2644 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2645 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2646 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2647 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2648 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2649 // CHECK9-NEXT: [[I6:%.*]] = alloca i32, align 4
2650 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2651 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2652 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2653 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2654 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2655 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2656 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2657 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2658 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2659 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2660 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2661 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2662 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2663 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2664 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2665 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2666 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2667 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2668 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2669 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2670 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2671 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2672 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2673 // CHECK9: omp.precond.then:
2674 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2675 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2676 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2677 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2678 // CHECK9-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
2679 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2680 // CHECK9-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
2681 // CHECK9-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
2682 // CHECK9-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
2683 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2684 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2685 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2686 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2687 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2688 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2689 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2690 // CHECK9-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2691 // CHECK9-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2692 // CHECK9: cond.true:
2693 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2694 // CHECK9-NEXT: br label [[COND_END:%.*]]
2695 // CHECK9: cond.false:
2696 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2697 // CHECK9-NEXT: br label [[COND_END]]
2698 // CHECK9: cond.end:
2699 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2700 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2701 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2702 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
2703 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2704 // CHECK9: omp.inner.for.cond:
2705 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2706 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
2707 // CHECK9-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2708 // CHECK9-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2709 // CHECK9: omp.inner.for.body:
2710 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2711 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2712 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2713 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !27
2714 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !27
2715 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2716 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
2717 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !27
2718 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2719 // CHECK9: omp.body.continue:
2720 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2721 // CHECK9: omp.inner.for.inc:
2722 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2723 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
2724 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
2725 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
2726 // CHECK9: omp.inner.for.end:
2727 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2728 // CHECK9: omp.loop.exit:
2729 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2730 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2731 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2732 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2733 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
2734 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2735 // CHECK9: .omp.final.then:
2736 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2737 // CHECK9-NEXT: [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
2738 // CHECK9-NEXT: [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
2739 // CHECK9-NEXT: [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
2740 // CHECK9-NEXT: [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
2741 // CHECK9-NEXT: store i32 [[ADD13]], i32* [[I6]], align 4
2742 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2743 // CHECK9: .omp.final.done:
2744 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2745 // CHECK9: omp.precond.end:
2746 // CHECK9-NEXT: ret void
2747 //
2748 //
2749 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2750 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
2751 // CHECK9-NEXT: entry:
2752 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2753 // CHECK9-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
2754 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
2755 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2756 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2757 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2758 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2759 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
2760 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
2761 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
2762 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
2763 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
2764 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
2765 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
2766 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
2767 // CHECK9-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
2768 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2769 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
2770 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2771 // CHECK9-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
2772 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
2773 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2774 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
2775 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
2776 // CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2777 // CHECK9-NEXT: store i8* null, i8** [[TMP4]], align 8
2778 // CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2779 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2780 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2781 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2782 // CHECK9-NEXT: store i32 1, i32* [[TMP7]], align 4
2783 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2784 // CHECK9-NEXT: store i32 1, i32* [[TMP8]], align 4
2785 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2786 // CHECK9-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
2787 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2788 // CHECK9-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
2789 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2790 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP11]], align 8
2791 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2792 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP12]], align 8
2793 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2794 // CHECK9-NEXT: store i8** null, i8*** [[TMP13]], align 8
2795 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2796 // CHECK9-NEXT: store i8** null, i8*** [[TMP14]], align 8
2797 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2798 // CHECK9-NEXT: store i64 10, i64* [[TMP15]], align 8
2799 // CHECK9-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2800 // CHECK9-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2801 // CHECK9-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2802 // CHECK9: omp_offload.failed:
2803 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
2804 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
2805 // CHECK9: omp_offload.cont:
2806 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2807 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
2808 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
2809 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2810 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
2811 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
2812 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
2813 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
2814 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2815 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2816 // CHECK9-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2817 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
2818 // CHECK9-NEXT: store i32 1, i32* [[TMP25]], align 4
2819 // CHECK9-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
2820 // CHECK9-NEXT: store i32 1, i32* [[TMP26]], align 4
2821 // CHECK9-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
2822 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
2823 // CHECK9-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
2824 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
2825 // CHECK9-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
2826 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP29]], align 8
2827 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
2828 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP30]], align 8
2829 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
2830 // CHECK9-NEXT: store i8** null, i8*** [[TMP31]], align 8
2831 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
2832 // CHECK9-NEXT: store i8** null, i8*** [[TMP32]], align 8
2833 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
2834 // CHECK9-NEXT: store i64 10, i64* [[TMP33]], align 8
2835 // CHECK9-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
2836 // CHECK9-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
2837 // CHECK9-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2838 // CHECK9: omp_offload.failed6:
2839 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
2840 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT7]]
2841 // CHECK9: omp_offload.cont7:
2842 // CHECK9-NEXT: [[TMP36:%.*]] = load i32, i32* [[M]], align 4
2843 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
2844 // CHECK9-NEXT: store i32 [[TMP36]], i32* [[CONV]], align 4
2845 // CHECK9-NEXT: [[TMP37:%.*]] = load i64, i64* [[M_CASTED]], align 8
2846 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2847 // CHECK9-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
2848 // CHECK9-NEXT: store i64 [[TMP37]], i64* [[TMP39]], align 8
2849 // CHECK9-NEXT: [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2850 // CHECK9-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
2851 // CHECK9-NEXT: store i64 [[TMP37]], i64* [[TMP41]], align 8
2852 // CHECK9-NEXT: [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
2853 // CHECK9-NEXT: store i8* null, i8** [[TMP42]], align 8
2854 // CHECK9-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
2855 // CHECK9-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
2856 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 8
2857 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
2858 // CHECK9-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
2859 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 8
2860 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
2861 // CHECK9-NEXT: store i8* null, i8** [[TMP47]], align 8
2862 // CHECK9-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2863 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2864 // CHECK9-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2865 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
2866 // CHECK9-NEXT: store i32 1, i32* [[TMP50]], align 4
2867 // CHECK9-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
2868 // CHECK9-NEXT: store i32 2, i32* [[TMP51]], align 4
2869 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
2870 // CHECK9-NEXT: store i8** [[TMP48]], i8*** [[TMP52]], align 8
2871 // CHECK9-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
2872 // CHECK9-NEXT: store i8** [[TMP49]], i8*** [[TMP53]], align 8
2873 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
2874 // CHECK9-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP54]], align 8
2875 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
2876 // CHECK9-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP55]], align 8
2877 // CHECK9-NEXT: [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
2878 // CHECK9-NEXT: store i8** null, i8*** [[TMP56]], align 8
2879 // CHECK9-NEXT: [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
2880 // CHECK9-NEXT: store i8** null, i8*** [[TMP57]], align 8
2881 // CHECK9-NEXT: [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
2882 // CHECK9-NEXT: store i64 10, i64* [[TMP58]], align 8
2883 // CHECK9-NEXT: [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
2884 // CHECK9-NEXT: [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
2885 // CHECK9-NEXT: br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2886 // CHECK9: omp_offload.failed13:
2887 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i64 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
2888 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT14]]
2889 // CHECK9: omp_offload.cont14:
2890 // CHECK9-NEXT: ret i32 0
2891 //
2892 //
2893 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
2894 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2895 // CHECK9-NEXT: entry:
2896 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2897 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2898 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2899 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2900 // CHECK9-NEXT: ret void
2901 //
2902 //
2903 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
2904 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2905 // CHECK9-NEXT: entry:
2906 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2907 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2908 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2909 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2910 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2911 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2912 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2913 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2914 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2915 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2916 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2917 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2918 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2919 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2920 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2921 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
2922 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2923 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2924 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2925 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2926 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2927 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2928 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2929 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2930 // CHECK9: cond.true:
2931 // CHECK9-NEXT: br label [[COND_END:%.*]]
2932 // CHECK9: cond.false:
2933 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2934 // CHECK9-NEXT: br label [[COND_END]]
2935 // CHECK9: cond.end:
2936 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2937 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2938 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2939 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2940 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2941 // CHECK9: omp.inner.for.cond:
2942 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
2943 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
2944 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2945 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2946 // CHECK9: omp.inner.for.body:
2947 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
2948 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2949 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
2950 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2951 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !30
2952 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2953 // CHECK9: omp.inner.for.inc:
2954 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
2955 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
2956 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2957 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
2958 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
2959 // CHECK9: omp.inner.for.end:
2960 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2961 // CHECK9: omp.loop.exit:
2962 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2963 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2964 // CHECK9-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2965 // CHECK9-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2966 // CHECK9: .omp.final.then:
2967 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
2968 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
2969 // CHECK9: .omp.final.done:
2970 // CHECK9-NEXT: ret void
2971 //
2972 //
2973 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
2974 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2975 // CHECK9-NEXT: entry:
2976 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2977 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2978 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2979 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2980 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2981 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2982 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2983 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2984 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2985 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2986 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2987 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2988 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2989 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2990 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2991 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2992 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2993 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2994 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2995 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2996 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2997 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
2998 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2999 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
3000 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3001 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3002 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3003 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3004 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3005 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3006 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3007 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3008 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
3009 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3010 // CHECK9: cond.true:
3011 // CHECK9-NEXT: br label [[COND_END:%.*]]
3012 // CHECK9: cond.false:
3013 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3014 // CHECK9-NEXT: br label [[COND_END]]
3015 // CHECK9: cond.end:
3016 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3017 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3018 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3019 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3020 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3021 // CHECK9: omp.inner.for.cond:
3022 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
3023 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
3024 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3025 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3026 // CHECK9: omp.inner.for.body:
3027 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
3028 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3029 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3030 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
3031 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !33
3032 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3033 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3034 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !33
3035 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3036 // CHECK9: omp.body.continue:
3037 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3038 // CHECK9: omp.inner.for.inc:
3039 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
3040 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
3041 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
3042 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
3043 // CHECK9: omp.inner.for.end:
3044 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3045 // CHECK9: omp.loop.exit:
3046 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3047 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3048 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3049 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3050 // CHECK9: .omp.final.then:
3051 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
3052 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3053 // CHECK9: .omp.final.done:
3054 // CHECK9-NEXT: ret void
3055 //
3056 //
3057 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
3058 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3059 // CHECK9-NEXT: entry:
3060 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3061 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3062 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3063 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
3064 // CHECK9-NEXT: ret void
3065 //
3066 //
3067 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..14
3068 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3069 // CHECK9-NEXT: entry:
3070 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3071 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3072 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3073 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3074 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3075 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3076 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3077 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3078 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3079 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3080 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3081 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3082 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3083 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3084 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3085 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
3086 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3087 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3088 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3089 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3090 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3091 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3092 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3093 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3094 // CHECK9: cond.true:
3095 // CHECK9-NEXT: br label [[COND_END:%.*]]
3096 // CHECK9: cond.false:
3097 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3098 // CHECK9-NEXT: br label [[COND_END]]
3099 // CHECK9: cond.end:
3100 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3101 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3102 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3103 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3104 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3105 // CHECK9: omp.inner.for.cond:
3106 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
3107 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
3108 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3109 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3110 // CHECK9: omp.inner.for.body:
3111 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
3112 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
3113 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
3114 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3115 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]]), !llvm.access.group !36
3116 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3117 // CHECK9: omp.inner.for.inc:
3118 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
3119 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
3120 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3121 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
3122 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
3123 // CHECK9: omp.inner.for.end:
3124 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3125 // CHECK9: omp.loop.exit:
3126 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3127 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3128 // CHECK9-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
3129 // CHECK9-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3130 // CHECK9: .omp.final.then:
3131 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
3132 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3133 // CHECK9: .omp.final.done:
3134 // CHECK9-NEXT: ret void
3135 //
3136 //
3137 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
3138 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3139 // CHECK9-NEXT: entry:
3140 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3141 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3142 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3143 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3144 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3145 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3146 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3147 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3148 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3149 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3150 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3151 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3152 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3153 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3154 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3155 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3156 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3157 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3158 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3159 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3160 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3161 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
3162 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3163 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
3164 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3165 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3166 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3167 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3168 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3169 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3170 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3171 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3172 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
3173 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3174 // CHECK9: cond.true:
3175 // CHECK9-NEXT: br label [[COND_END:%.*]]
3176 // CHECK9: cond.false:
3177 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3178 // CHECK9-NEXT: br label [[COND_END]]
3179 // CHECK9: cond.end:
3180 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3181 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3182 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3183 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3184 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3185 // CHECK9: omp.inner.for.cond:
3186 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3187 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
3188 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3189 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3190 // CHECK9: omp.inner.for.body:
3191 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3192 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3193 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3194 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
3195 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !39
3196 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3197 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3198 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !39
3199 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3200 // CHECK9: omp.body.continue:
3201 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3202 // CHECK9: omp.inner.for.inc:
3203 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3204 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
3205 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3206 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
3207 // CHECK9: omp.inner.for.end:
3208 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3209 // CHECK9: omp.loop.exit:
3210 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3211 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3212 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3213 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3214 // CHECK9: .omp.final.then:
3215 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
3216 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3217 // CHECK9: .omp.final.done:
3218 // CHECK9-NEXT: ret void
3219 //
3220 //
3221 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
3222 // CHECK9-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3223 // CHECK9-NEXT: entry:
3224 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
3225 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3226 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3227 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3228 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
3229 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3230 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
3231 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3232 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
3233 // CHECK9-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
3234 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3235 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3236 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
3237 // CHECK9-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3238 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
3239 // CHECK9-NEXT: ret void
3240 //
3241 //
3242 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..18
3243 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
3244 // CHECK9-NEXT: entry:
3245 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3246 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3247 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3248 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3249 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3250 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3251 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3252 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3253 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3254 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3255 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3256 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3257 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3258 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3259 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3260 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3261 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3262 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3263 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3264 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
3265 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3266 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3267 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
3268 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3269 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3270 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
3271 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3272 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
3273 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3274 // CHECK9: cond.true:
3275 // CHECK9-NEXT: br label [[COND_END:%.*]]
3276 // CHECK9: cond.false:
3277 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3278 // CHECK9-NEXT: br label [[COND_END]]
3279 // CHECK9: cond.end:
3280 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3281 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3282 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3283 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3284 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3285 // CHECK9: omp.inner.for.cond:
3286 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
3287 // CHECK9-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
3288 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3289 // CHECK9: omp.inner.for.body:
3290 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
3291 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
3292 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3293 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3294 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4, !llvm.access.group !42
3295 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3296 // CHECK9-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4, !llvm.access.group !42
3297 // CHECK9-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !42
3298 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]]), !llvm.access.group !42
3299 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3300 // CHECK9: omp.inner.for.inc:
3301 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
3302 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
3303 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
3304 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
3305 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
3306 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
3307 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
3308 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
3309 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3310 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
3311 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3312 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3313 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3314 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP20]], 9
3315 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
3316 // CHECK9: cond.true6:
3317 // CHECK9-NEXT: br label [[COND_END8:%.*]]
3318 // CHECK9: cond.false7:
3319 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3320 // CHECK9-NEXT: br label [[COND_END8]]
3321 // CHECK9: cond.end8:
3322 // CHECK9-NEXT: [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP21]], [[COND_FALSE7]] ]
3323 // CHECK9-NEXT: store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
3324 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
3325 // CHECK9-NEXT: store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
3326 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
3327 // CHECK9: omp.inner.for.end:
3328 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3329 // CHECK9: omp.loop.exit:
3330 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3331 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3332 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3333 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3334 // CHECK9: .omp.final.then:
3335 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
3336 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3337 // CHECK9: .omp.final.done:
3338 // CHECK9-NEXT: ret void
3339 //
3340 //
3341 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..19
3342 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
3343 // CHECK9-NEXT: entry:
3344 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3345 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3346 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3347 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3348 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3349 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3350 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3351 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3352 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3353 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3354 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3355 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3356 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3357 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3358 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3359 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3360 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3361 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3362 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3363 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3364 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3365 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3366 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3367 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3368 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3369 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3370 // CHECK9-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
3371 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
3372 // CHECK9-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
3373 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3374 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3375 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3376 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3377 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3378 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3379 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
3380 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3381 // CHECK9: cond.true:
3382 // CHECK9-NEXT: br label [[COND_END:%.*]]
3383 // CHECK9: cond.false:
3384 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3385 // CHECK9-NEXT: br label [[COND_END]]
3386 // CHECK9: cond.end:
3387 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3388 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3389 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3390 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3391 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3392 // CHECK9: omp.inner.for.cond:
3393 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
3394 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !45
3395 // CHECK9-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3396 // CHECK9-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3397 // CHECK9: omp.inner.for.body:
3398 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
3399 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3400 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3401 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !45
3402 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !45
3403 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
3404 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3405 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !45
3406 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3407 // CHECK9: omp.body.continue:
3408 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3409 // CHECK9: omp.inner.for.inc:
3410 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
3411 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
3412 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
3413 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
3414 // CHECK9: omp.inner.for.end:
3415 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3416 // CHECK9: omp.loop.exit:
3417 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3418 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3419 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3420 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3421 // CHECK9: .omp.final.then:
3422 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
3423 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3424 // CHECK9: .omp.final.done:
3425 // CHECK9-NEXT: ret void
3426 //
3427 //
3428 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3429 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
3430 // CHECK9-NEXT: entry:
3431 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
3432 // CHECK9-NEXT: ret void
3433 //
3434 //
3435 // CHECK11-LABEL: define {{[^@]+}}@main
3436 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3437 // CHECK11-NEXT: entry:
3438 // CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3439 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3440 // CHECK11-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
3441 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
3442 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
3443 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3444 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
3445 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
3446 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3447 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3448 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3449 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
3450 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3451 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3452 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3453 // CHECK11-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
3454 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
3455 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
3456 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
3457 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
3458 // CHECK11-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
3459 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
3460 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
3461 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
3462 // CHECK11-NEXT: [[N_CASTED18:%.*]] = alloca i32, align 4
3463 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [4 x i8*], align 4
3464 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS20:%.*]] = alloca [4 x i8*], align 4
3465 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [4 x i8*], align 4
3466 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES22:%.*]] = alloca [4 x i64], align 4
3467 // CHECK11-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
3468 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
3469 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
3470 // CHECK11-NEXT: store i32 0, i32* [[RETVAL]], align 4
3471 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3472 // CHECK11-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
3473 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
3474 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3475 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
3476 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
3477 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
3478 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
3479 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
3480 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
3481 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
3482 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
3483 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
3484 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
3485 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3486 // 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)
3487 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3488 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3489 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
3490 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3491 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3492 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
3493 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3494 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
3495 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3496 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3497 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
3498 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3499 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
3500 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
3501 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3502 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
3503 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3504 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
3505 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
3506 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3507 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
3508 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
3509 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3510 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
3511 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3512 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
3513 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3514 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3515 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3516 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
3517 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
3518 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3519 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
3520 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3521 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3522 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3523 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3524 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
3525 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
3526 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3527 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3528 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4
3529 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3530 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4
3531 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3532 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
3533 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3534 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
3535 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3536 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
3537 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3538 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
3539 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3540 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4
3541 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3542 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4
3543 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3544 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
3545 // CHECK11-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3546 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
3547 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3548 // CHECK11: omp_offload.failed:
3549 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3550 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
3551 // CHECK11: omp_offload.cont:
3552 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
3553 // CHECK11-NEXT: store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
3554 // CHECK11-NEXT: [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
3555 // CHECK11-NEXT: [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
3556 // CHECK11-NEXT: [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
3557 // CHECK11-NEXT: [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
3558 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i32 24, i1 false)
3559 // CHECK11-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3560 // CHECK11-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
3561 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP47]], align 4
3562 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3563 // CHECK11-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
3564 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP49]], align 4
3565 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
3566 // CHECK11-NEXT: store i8* null, i8** [[TMP50]], align 4
3567 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
3568 // CHECK11-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
3569 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP52]], align 4
3570 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
3571 // CHECK11-NEXT: [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
3572 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP54]], align 4
3573 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
3574 // CHECK11-NEXT: store i8* null, i8** [[TMP55]], align 4
3575 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
3576 // CHECK11-NEXT: [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
3577 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP57]], align 4
3578 // CHECK11-NEXT: [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
3579 // CHECK11-NEXT: [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
3580 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP59]], align 4
3581 // CHECK11-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
3582 // CHECK11-NEXT: store i64 [[TMP44]], i64* [[TMP60]], align 4
3583 // CHECK11-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
3584 // CHECK11-NEXT: store i8* null, i8** [[TMP61]], align 4
3585 // CHECK11-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3586 // CHECK11-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3587 // CHECK11-NEXT: [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
3588 // CHECK11-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
3589 // CHECK11-NEXT: store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
3590 // CHECK11-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
3591 // CHECK11-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 0
3592 // CHECK11-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
3593 // CHECK11-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
3594 // CHECK11-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
3595 // CHECK11-NEXT: [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
3596 // CHECK11-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
3597 // CHECK11-NEXT: [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
3598 // CHECK11-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
3599 // CHECK11-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
3600 // CHECK11-NEXT: store i32 1, i32* [[TMP69]], align 4
3601 // CHECK11-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
3602 // CHECK11-NEXT: store i32 3, i32* [[TMP70]], align 4
3603 // CHECK11-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
3604 // CHECK11-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 4
3605 // CHECK11-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
3606 // CHECK11-NEXT: store i8** [[TMP63]], i8*** [[TMP72]], align 4
3607 // CHECK11-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
3608 // CHECK11-NEXT: store i64* [[TMP64]], i64** [[TMP73]], align 4
3609 // CHECK11-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
3610 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP74]], align 4
3611 // CHECK11-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
3612 // CHECK11-NEXT: store i8** null, i8*** [[TMP75]], align 4
3613 // CHECK11-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
3614 // CHECK11-NEXT: store i8** null, i8*** [[TMP76]], align 4
3615 // CHECK11-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
3616 // CHECK11-NEXT: store i64 [[TMP68]], i64* [[TMP77]], align 8
3617 // CHECK11-NEXT: [[TMP78:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
3618 // CHECK11-NEXT: [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
3619 // CHECK11-NEXT: br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
3620 // CHECK11: omp_offload.failed16:
3621 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
3622 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT17]]
3623 // CHECK11: omp_offload.cont17:
3624 // CHECK11-NEXT: [[TMP80:%.*]] = load i32, i32* [[M]], align 4
3625 // CHECK11-NEXT: store i32 [[TMP80]], i32* [[M_CASTED]], align 4
3626 // CHECK11-NEXT: [[TMP81:%.*]] = load i32, i32* [[M_CASTED]], align 4
3627 // CHECK11-NEXT: [[TMP82:%.*]] = load i32, i32* [[N]], align 4
3628 // CHECK11-NEXT: store i32 [[TMP82]], i32* [[N_CASTED18]], align 4
3629 // CHECK11-NEXT: [[TMP83:%.*]] = load i32, i32* [[N_CASTED18]], align 4
3630 // CHECK11-NEXT: [[TMP84:%.*]] = mul nuw i32 [[TMP0]], 4
3631 // CHECK11-NEXT: [[TMP85:%.*]] = sext i32 [[TMP84]] to i64
3632 // CHECK11-NEXT: [[TMP86:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
3633 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP86]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i32 32, i1 false)
3634 // CHECK11-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
3635 // CHECK11-NEXT: [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
3636 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP88]], align 4
3637 // CHECK11-NEXT: [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
3638 // CHECK11-NEXT: [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
3639 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP90]], align 4
3640 // CHECK11-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
3641 // CHECK11-NEXT: store i8* null, i8** [[TMP91]], align 4
3642 // CHECK11-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
3643 // CHECK11-NEXT: [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
3644 // CHECK11-NEXT: store i32 [[TMP83]], i32* [[TMP93]], align 4
3645 // CHECK11-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
3646 // CHECK11-NEXT: [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
3647 // CHECK11-NEXT: store i32 [[TMP83]], i32* [[TMP95]], align 4
3648 // CHECK11-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
3649 // CHECK11-NEXT: store i8* null, i8** [[TMP96]], align 4
3650 // CHECK11-NEXT: [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
3651 // CHECK11-NEXT: [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32*
3652 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP98]], align 4
3653 // CHECK11-NEXT: [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
3654 // CHECK11-NEXT: [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
3655 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP100]], align 4
3656 // CHECK11-NEXT: [[TMP101:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
3657 // CHECK11-NEXT: store i8* null, i8** [[TMP101]], align 4
3658 // CHECK11-NEXT: [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 3
3659 // CHECK11-NEXT: [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
3660 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP103]], align 4
3661 // CHECK11-NEXT: [[TMP104:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 3
3662 // CHECK11-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i32**
3663 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP105]], align 4
3664 // CHECK11-NEXT: [[TMP106:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 3
3665 // CHECK11-NEXT: store i64 [[TMP85]], i64* [[TMP106]], align 4
3666 // CHECK11-NEXT: [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 3
3667 // CHECK11-NEXT: store i8* null, i8** [[TMP107]], align 4
3668 // CHECK11-NEXT: [[TMP108:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
3669 // CHECK11-NEXT: [[TMP109:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
3670 // CHECK11-NEXT: [[TMP110:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
3671 // CHECK11-NEXT: [[TMP111:%.*]] = load i32, i32* [[N]], align 4
3672 // CHECK11-NEXT: store i32 [[TMP111]], i32* [[DOTCAPTURE_EXPR_24]], align 4
3673 // CHECK11-NEXT: [[TMP112:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
3674 // CHECK11-NEXT: [[SUB26:%.*]] = sub nsw i32 [[TMP112]], 0
3675 // CHECK11-NEXT: [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
3676 // CHECK11-NEXT: [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
3677 // CHECK11-NEXT: store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
3678 // CHECK11-NEXT: [[TMP113:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
3679 // CHECK11-NEXT: [[ADD29:%.*]] = add nsw i32 [[TMP113]], 1
3680 // CHECK11-NEXT: [[TMP114:%.*]] = zext i32 [[ADD29]] to i64
3681 // CHECK11-NEXT: [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
3682 // CHECK11-NEXT: [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
3683 // CHECK11-NEXT: store i32 1, i32* [[TMP115]], align 4
3684 // CHECK11-NEXT: [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
3685 // CHECK11-NEXT: store i32 4, i32* [[TMP116]], align 4
3686 // CHECK11-NEXT: [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
3687 // CHECK11-NEXT: store i8** [[TMP108]], i8*** [[TMP117]], align 4
3688 // CHECK11-NEXT: [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
3689 // CHECK11-NEXT: store i8** [[TMP109]], i8*** [[TMP118]], align 4
3690 // CHECK11-NEXT: [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
3691 // CHECK11-NEXT: store i64* [[TMP110]], i64** [[TMP119]], align 4
3692 // CHECK11-NEXT: [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
3693 // CHECK11-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP120]], align 4
3694 // CHECK11-NEXT: [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
3695 // CHECK11-NEXT: store i8** null, i8*** [[TMP121]], align 4
3696 // CHECK11-NEXT: [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
3697 // CHECK11-NEXT: store i8** null, i8*** [[TMP122]], align 4
3698 // CHECK11-NEXT: [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
3699 // CHECK11-NEXT: store i64 [[TMP114]], i64* [[TMP123]], align 8
3700 // CHECK11-NEXT: [[TMP124:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
3701 // CHECK11-NEXT: [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
3702 // CHECK11-NEXT: br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
3703 // CHECK11: omp_offload.failed31:
3704 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i32 [[TMP81]], i32 [[TMP83]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
3705 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT32]]
3706 // CHECK11: omp_offload.cont32:
3707 // CHECK11-NEXT: [[TMP126:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3708 // CHECK11-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP126]])
3709 // CHECK11-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3710 // CHECK11-NEXT: [[TMP127:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3711 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP127]])
3712 // CHECK11-NEXT: [[TMP128:%.*]] = load i32, i32* [[RETVAL]], align 4
3713 // CHECK11-NEXT: ret i32 [[TMP128]]
3714 //
3715 //
3716 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
3717 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3718 // CHECK11-NEXT: entry:
3719 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3720 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3721 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3722 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3723 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3724 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3725 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3726 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3727 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], 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]])
3728 // CHECK11-NEXT: ret void
3729 //
3730 //
3731 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
3732 // 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]] {
3733 // CHECK11-NEXT: entry:
3734 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3735 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3736 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3737 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3738 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3739 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3740 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3741 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3742 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3743 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3744 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3745 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3746 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3747 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3748 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3749 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3750 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3751 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3752 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3753 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3754 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3755 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3756 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3757 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3758 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3759 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3760 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3761 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3762 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3763 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3764 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3765 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3766 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3767 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3768 // CHECK11: omp.precond.then:
3769 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3770 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3771 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3772 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3773 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3774 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3775 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3776 // 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_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3777 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3778 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3779 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3780 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3781 // CHECK11: cond.true:
3782 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3783 // CHECK11-NEXT: br label [[COND_END:%.*]]
3784 // CHECK11: cond.false:
3785 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3786 // CHECK11-NEXT: br label [[COND_END]]
3787 // CHECK11: cond.end:
3788 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3789 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3790 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3791 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3792 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3793 // CHECK11: omp.inner.for.cond:
3794 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3795 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !10
3796 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3797 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3798 // CHECK11: omp.inner.for.body:
3799 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !10
3800 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !10
3801 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !10
3802 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3803 // CHECK11: omp.inner.for.inc:
3804 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3805 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !10
3806 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3807 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
3808 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
3809 // CHECK11: omp.inner.for.end:
3810 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3811 // CHECK11: omp.loop.exit:
3812 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3813 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
3814 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
3815 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3816 // CHECK11-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
3817 // CHECK11-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3818 // CHECK11: .omp.final.then:
3819 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3820 // CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
3821 // CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
3822 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
3823 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
3824 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
3825 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
3826 // CHECK11: .omp.final.done:
3827 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3828 // CHECK11: omp.precond.end:
3829 // CHECK11-NEXT: ret void
3830 //
3831 //
3832 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
3833 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3834 // CHECK11-NEXT: entry:
3835 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3836 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3837 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3838 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3839 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3840 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3841 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3842 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3843 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3844 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3845 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3846 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3847 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3848 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3849 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3850 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3851 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3852 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3853 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3854 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3855 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3856 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3857 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3858 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3859 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3860 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3861 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3862 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3863 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3864 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3865 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3866 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3867 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3868 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3869 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3870 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3871 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3872 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3873 // CHECK11: omp.precond.then:
3874 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3875 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3876 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3877 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3878 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3879 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
3880 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
3881 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3882 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3883 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3884 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3885 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3886 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3887 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3888 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3889 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3890 // CHECK11: cond.true:
3891 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3892 // CHECK11-NEXT: br label [[COND_END:%.*]]
3893 // CHECK11: cond.false:
3894 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3895 // CHECK11-NEXT: br label [[COND_END]]
3896 // CHECK11: cond.end:
3897 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3898 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3899 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3900 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3901 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3902 // CHECK11: omp.inner.for.cond:
3903 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3904 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
3905 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3906 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3907 // CHECK11: omp.inner.for.body:
3908 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3909 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3910 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3911 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !14
3912 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !14
3913 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
3914 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
3915 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3916 // CHECK11: omp.body.continue:
3917 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3918 // CHECK11: omp.inner.for.inc:
3919 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3920 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
3921 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3922 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
3923 // CHECK11: omp.inner.for.end:
3924 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3925 // CHECK11: omp.loop.exit:
3926 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3927 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3928 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3929 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3930 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3931 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3932 // CHECK11: .omp.final.then:
3933 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3934 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
3935 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
3936 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
3937 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
3938 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
3939 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
3940 // CHECK11: .omp.final.done:
3941 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3942 // CHECK11: omp.precond.end:
3943 // CHECK11-NEXT: ret void
3944 //
3945 //
3946 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
3947 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3948 // CHECK11-NEXT: entry:
3949 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3950 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3951 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3952 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3953 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3954 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3955 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3956 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3957 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
3958 // CHECK11-NEXT: ret void
3959 //
3960 //
3961 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
3962 // 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]] {
3963 // CHECK11-NEXT: entry:
3964 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3965 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3966 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3967 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3968 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3969 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3970 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3971 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3972 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3973 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3974 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3975 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3976 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3977 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3978 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3979 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3980 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3981 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3982 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3983 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3984 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3985 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3986 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3987 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3988 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3989 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3990 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3991 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3992 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3993 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3994 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3995 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3996 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3997 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3998 // CHECK11: omp.precond.then:
3999 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4000 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4001 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
4002 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4003 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4004 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4005 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4006 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4007 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4008 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4009 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
4010 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4011 // CHECK11: cond.true:
4012 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4013 // CHECK11-NEXT: br label [[COND_END:%.*]]
4014 // CHECK11: cond.false:
4015 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4016 // CHECK11-NEXT: br label [[COND_END]]
4017 // CHECK11: cond.end:
4018 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4019 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4020 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4021 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4022 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4023 // CHECK11: omp.inner.for.cond:
4024 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
4025 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
4026 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
4027 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4028 // CHECK11: omp.inner.for.body:
4029 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !19
4030 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !19
4031 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]]), !llvm.access.group !19
4032 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4033 // CHECK11: omp.inner.for.inc:
4034 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
4035 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !19
4036 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
4037 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
4038 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
4039 // CHECK11: omp.inner.for.end:
4040 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4041 // CHECK11: omp.loop.exit:
4042 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4043 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
4044 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
4045 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4046 // CHECK11-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
4047 // CHECK11-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4048 // CHECK11: .omp.final.then:
4049 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4050 // CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
4051 // CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
4052 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
4053 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
4054 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
4055 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4056 // CHECK11: .omp.final.done:
4057 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
4058 // CHECK11: omp.precond.end:
4059 // CHECK11-NEXT: ret void
4060 //
4061 //
4062 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
4063 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
4064 // CHECK11-NEXT: entry:
4065 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4066 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4067 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4068 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4069 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
4070 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4071 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4072 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4073 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4074 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4075 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4076 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4077 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4078 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4079 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4080 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4081 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
4082 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4083 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4084 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4085 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4086 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
4087 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4088 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4089 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
4090 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4091 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4092 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4093 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
4094 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4095 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4096 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4097 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4098 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4099 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
4100 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4101 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4102 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4103 // CHECK11: omp.precond.then:
4104 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4105 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4106 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
4107 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4108 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4109 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
4110 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
4111 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4112 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4113 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4114 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4115 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4116 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4117 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4118 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
4119 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4120 // CHECK11: cond.true:
4121 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4122 // CHECK11-NEXT: br label [[COND_END:%.*]]
4123 // CHECK11: cond.false:
4124 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4125 // CHECK11-NEXT: br label [[COND_END]]
4126 // CHECK11: cond.end:
4127 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
4128 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4129 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4130 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
4131 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4132 // CHECK11: omp.inner.for.cond:
4133 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
4134 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
4135 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
4136 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4137 // CHECK11: omp.inner.for.body:
4138 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
4139 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
4140 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4141 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !22
4142 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !22
4143 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
4144 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
4145 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4146 // CHECK11: omp.body.continue:
4147 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4148 // CHECK11: omp.inner.for.inc:
4149 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
4150 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
4151 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
4152 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
4153 // CHECK11: omp.inner.for.end:
4154 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4155 // CHECK11: omp.loop.exit:
4156 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4157 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
4158 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
4159 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4160 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4161 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4162 // CHECK11: .omp.final.then:
4163 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4164 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
4165 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
4166 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
4167 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
4168 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
4169 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4170 // CHECK11: .omp.final.done:
4171 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
4172 // CHECK11: omp.precond.end:
4173 // CHECK11-NEXT: ret void
4174 //
4175 //
4176 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
4177 // CHECK11-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
4178 // CHECK11-NEXT: entry:
4179 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
4180 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
4181 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4182 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4183 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4184 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4185 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
4186 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
4187 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4188 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4189 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4190 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4191 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
4192 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
4193 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4194 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4195 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4196 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
4197 // CHECK11-NEXT: ret void
4198 //
4199 //
4200 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
4201 // 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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4202 // CHECK11-NEXT: entry:
4203 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4204 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4205 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
4206 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4207 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4208 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4209 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4210 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4211 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4212 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4213 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4214 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4215 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4216 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4217 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4218 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
4219 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4220 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4221 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4222 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
4223 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4224 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4225 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4226 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
4227 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4228 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4229 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4230 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4231 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4232 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4233 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4234 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
4235 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4236 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
4237 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4238 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4239 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4240 // CHECK11: omp.precond.then:
4241 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4242 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4243 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
4244 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4245 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4246 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4247 // CHECK11-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4248 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
4249 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
4250 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4251 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4252 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
4253 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4254 // CHECK11: cond.true:
4255 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4256 // CHECK11-NEXT: br label [[COND_END:%.*]]
4257 // CHECK11: cond.false:
4258 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4259 // CHECK11-NEXT: br label [[COND_END]]
4260 // CHECK11: cond.end:
4261 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
4262 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4263 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4264 // CHECK11-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
4265 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4266 // CHECK11: omp.inner.for.cond:
4267 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
4268 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !25
4269 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
4270 // CHECK11-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
4271 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4272 // CHECK11: omp.inner.for.body:
4273 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
4274 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4275 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !25
4276 // CHECK11-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !25
4277 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !25
4278 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]]), !llvm.access.group !25
4279 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4280 // CHECK11: omp.inner.for.inc:
4281 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
4282 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
4283 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
4284 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
4285 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
4286 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
4287 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
4288 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
4289 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4290 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !25
4291 // CHECK11-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
4292 // CHECK11-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4293 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4294 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !25
4295 // CHECK11-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
4296 // CHECK11-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
4297 // CHECK11: cond.true11:
4298 // CHECK11-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !25
4299 // CHECK11-NEXT: br label [[COND_END13:%.*]]
4300 // CHECK11: cond.false12:
4301 // CHECK11-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4302 // CHECK11-NEXT: br label [[COND_END13]]
4303 // CHECK11: cond.end13:
4304 // CHECK11-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
4305 // CHECK11-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !25
4306 // CHECK11-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !25
4307 // CHECK11-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
4308 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
4309 // CHECK11: omp.inner.for.end:
4310 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4311 // CHECK11: omp.loop.exit:
4312 // CHECK11-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4313 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
4314 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
4315 // CHECK11-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4316 // CHECK11-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
4317 // CHECK11-NEXT: br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4318 // CHECK11: .omp.final.then:
4319 // CHECK11-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4320 // CHECK11-NEXT: [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
4321 // CHECK11-NEXT: [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
4322 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
4323 // CHECK11-NEXT: [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
4324 // CHECK11-NEXT: store i32 [[ADD17]], i32* [[I4]], align 4
4325 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4326 // CHECK11: .omp.final.done:
4327 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
4328 // CHECK11: omp.precond.end:
4329 // CHECK11-NEXT: ret void
4330 //
4331 //
4332 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
4333 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4334 // CHECK11-NEXT: entry:
4335 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4336 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4337 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4338 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4339 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
4340 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4341 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4342 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4343 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4344 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4345 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4346 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
4347 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4348 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4349 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4350 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4351 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4352 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
4353 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4354 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4355 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4356 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4357 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
4358 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4359 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4360 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4361 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
4362 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4363 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4364 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4365 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4366 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4367 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4368 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4369 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
4370 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
4371 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
4372 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4373 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4374 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4375 // CHECK11: omp.precond.then:
4376 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4377 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4378 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
4379 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4380 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4381 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
4382 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
4383 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4384 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4385 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4386 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4387 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4388 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4389 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4390 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
4391 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4392 // CHECK11: cond.true:
4393 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
4394 // CHECK11-NEXT: br label [[COND_END:%.*]]
4395 // CHECK11: cond.false:
4396 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4397 // CHECK11-NEXT: br label [[COND_END]]
4398 // CHECK11: cond.end:
4399 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
4400 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4401 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4402 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
4403 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4404 // CHECK11: omp.inner.for.cond:
4405 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
4406 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !28
4407 // CHECK11-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
4408 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4409 // CHECK11: omp.inner.for.body:
4410 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
4411 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
4412 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4413 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !28
4414 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !28
4415 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
4416 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !28
4417 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4418 // CHECK11: omp.body.continue:
4419 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4420 // CHECK11: omp.inner.for.inc:
4421 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
4422 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
4423 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !28
4424 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]]
4425 // CHECK11: omp.inner.for.end:
4426 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4427 // CHECK11: omp.loop.exit:
4428 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4429 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
4430 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
4431 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4432 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4433 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4434 // CHECK11: .omp.final.then:
4435 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4436 // CHECK11-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
4437 // CHECK11-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
4438 // CHECK11-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
4439 // CHECK11-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
4440 // CHECK11-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
4441 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4442 // CHECK11: .omp.final.done:
4443 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
4444 // CHECK11: omp.precond.end:
4445 // CHECK11-NEXT: ret void
4446 //
4447 //
4448 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4449 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
4450 // CHECK11-NEXT: entry:
4451 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4452 // CHECK11-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4453 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
4454 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
4455 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
4456 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
4457 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4458 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
4459 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
4460 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
4461 // CHECK11-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
4462 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
4463 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 4
4464 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 4
4465 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 4
4466 // CHECK11-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
4467 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4468 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
4469 // CHECK11-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4470 // CHECK11-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
4471 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
4472 // CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4473 // CHECK11-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
4474 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
4475 // CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4476 // CHECK11-NEXT: store i8* null, i8** [[TMP4]], align 4
4477 // CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4478 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4479 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
4480 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
4481 // CHECK11-NEXT: store i32 1, i32* [[TMP7]], align 4
4482 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
4483 // CHECK11-NEXT: store i32 1, i32* [[TMP8]], align 4
4484 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
4485 // CHECK11-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
4486 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
4487 // CHECK11-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
4488 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
4489 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP11]], align 4
4490 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
4491 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP12]], align 4
4492 // CHECK11-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
4493 // CHECK11-NEXT: store i8** null, i8*** [[TMP13]], align 4
4494 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
4495 // CHECK11-NEXT: store i8** null, i8*** [[TMP14]], align 4
4496 // CHECK11-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
4497 // CHECK11-NEXT: store i64 10, i64* [[TMP15]], align 8
4498 // CHECK11-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
4499 // CHECK11-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
4500 // CHECK11-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4501 // CHECK11: omp_offload.failed:
4502 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
4503 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
4504 // CHECK11: omp_offload.cont:
4505 // CHECK11-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4506 // CHECK11-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
4507 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 4
4508 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4509 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
4510 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 4
4511 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
4512 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
4513 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4514 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4515 // CHECK11-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4516 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
4517 // CHECK11-NEXT: store i32 1, i32* [[TMP25]], align 4
4518 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
4519 // CHECK11-NEXT: store i32 1, i32* [[TMP26]], align 4
4520 // CHECK11-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
4521 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
4522 // CHECK11-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
4523 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
4524 // CHECK11-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
4525 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP29]], align 4
4526 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
4527 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP30]], align 4
4528 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
4529 // CHECK11-NEXT: store i8** null, i8*** [[TMP31]], align 4
4530 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
4531 // CHECK11-NEXT: store i8** null, i8*** [[TMP32]], align 4
4532 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
4533 // CHECK11-NEXT: store i64 10, i64* [[TMP33]], align 8
4534 // CHECK11-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
4535 // CHECK11-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
4536 // CHECK11-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4537 // CHECK11: omp_offload.failed6:
4538 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
4539 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT7]]
4540 // CHECK11: omp_offload.cont7:
4541 // CHECK11-NEXT: [[TMP36:%.*]] = load i32, i32* [[M]], align 4
4542 // CHECK11-NEXT: store i32 [[TMP36]], i32* [[M_CASTED]], align 4
4543 // CHECK11-NEXT: [[TMP37:%.*]] = load i32, i32* [[M_CASTED]], align 4
4544 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4545 // CHECK11-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i32*
4546 // CHECK11-NEXT: store i32 [[TMP37]], i32* [[TMP39]], align 4
4547 // CHECK11-NEXT: [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4548 // CHECK11-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
4549 // CHECK11-NEXT: store i32 [[TMP37]], i32* [[TMP41]], align 4
4550 // CHECK11-NEXT: [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
4551 // CHECK11-NEXT: store i8* null, i8** [[TMP42]], align 4
4552 // CHECK11-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
4553 // CHECK11-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
4554 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 4
4555 // CHECK11-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
4556 // CHECK11-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
4557 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 4
4558 // CHECK11-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
4559 // CHECK11-NEXT: store i8* null, i8** [[TMP47]], align 4
4560 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4561 // CHECK11-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4562 // CHECK11-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4563 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
4564 // CHECK11-NEXT: store i32 1, i32* [[TMP50]], align 4
4565 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
4566 // CHECK11-NEXT: store i32 2, i32* [[TMP51]], align 4
4567 // CHECK11-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
4568 // CHECK11-NEXT: store i8** [[TMP48]], i8*** [[TMP52]], align 4
4569 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
4570 // CHECK11-NEXT: store i8** [[TMP49]], i8*** [[TMP53]], align 4
4571 // CHECK11-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
4572 // CHECK11-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP54]], align 4
4573 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
4574 // CHECK11-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP55]], align 4
4575 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
4576 // CHECK11-NEXT: store i8** null, i8*** [[TMP56]], align 4
4577 // CHECK11-NEXT: [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
4578 // CHECK11-NEXT: store i8** null, i8*** [[TMP57]], align 4
4579 // CHECK11-NEXT: [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
4580 // CHECK11-NEXT: store i64 10, i64* [[TMP58]], align 8
4581 // CHECK11-NEXT: [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
4582 // CHECK11-NEXT: [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
4583 // CHECK11-NEXT: br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
4584 // CHECK11: omp_offload.failed13:
4585 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i32 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
4586 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT14]]
4587 // CHECK11: omp_offload.cont14:
4588 // CHECK11-NEXT: ret i32 0
4589 //
4590 //
4591 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
4592 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4593 // CHECK11-NEXT: entry:
4594 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4595 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4596 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4597 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4598 // CHECK11-NEXT: ret void
4599 //
4600 //
4601 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
4602 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4603 // CHECK11-NEXT: entry:
4604 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4605 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4606 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4607 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4608 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4609 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4610 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4611 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4612 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4613 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4614 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4615 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4616 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4617 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4618 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4619 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4620 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4621 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4622 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4623 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4624 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4625 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4626 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4627 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4628 // CHECK11: cond.true:
4629 // CHECK11-NEXT: br label [[COND_END:%.*]]
4630 // CHECK11: cond.false:
4631 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4632 // CHECK11-NEXT: br label [[COND_END]]
4633 // CHECK11: cond.end:
4634 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4635 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4636 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4637 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4638 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4639 // CHECK11: omp.inner.for.cond:
4640 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
4641 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
4642 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4643 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4644 // CHECK11: omp.inner.for.body:
4645 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !31
4646 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !31
4647 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !31
4648 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4649 // CHECK11: omp.inner.for.inc:
4650 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
4651 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !31
4652 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
4653 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !31
4654 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
4655 // CHECK11: omp.inner.for.end:
4656 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4657 // CHECK11: omp.loop.exit:
4658 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4659 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4660 // CHECK11-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
4661 // CHECK11-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4662 // CHECK11: .omp.final.then:
4663 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
4664 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4665 // CHECK11: .omp.final.done:
4666 // CHECK11-NEXT: ret void
4667 //
4668 //
4669 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
4670 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4671 // CHECK11-NEXT: entry:
4672 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4673 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4674 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4675 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4676 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4677 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4678 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4679 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4680 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4681 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4682 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4683 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4684 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4685 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4686 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4687 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4688 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4689 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4690 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4691 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4692 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4693 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4694 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
4695 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
4696 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4697 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4698 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4699 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4700 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4701 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4702 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4703 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4704 // CHECK11: cond.true:
4705 // CHECK11-NEXT: br label [[COND_END:%.*]]
4706 // CHECK11: cond.false:
4707 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4708 // CHECK11-NEXT: br label [[COND_END]]
4709 // CHECK11: cond.end:
4710 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4711 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4712 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4713 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4714 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4715 // CHECK11: omp.inner.for.cond:
4716 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
4717 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !34
4718 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4719 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4720 // CHECK11: omp.inner.for.body:
4721 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
4722 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4723 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4724 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !34
4725 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !34
4726 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
4727 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !34
4728 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4729 // CHECK11: omp.body.continue:
4730 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4731 // CHECK11: omp.inner.for.inc:
4732 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
4733 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
4734 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !34
4735 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP35:![0-9]+]]
4736 // CHECK11: omp.inner.for.end:
4737 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4738 // CHECK11: omp.loop.exit:
4739 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4740 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4741 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4742 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4743 // CHECK11: .omp.final.then:
4744 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
4745 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4746 // CHECK11: .omp.final.done:
4747 // CHECK11-NEXT: ret void
4748 //
4749 //
4750 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
4751 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4752 // CHECK11-NEXT: entry:
4753 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4754 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4755 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4756 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4757 // CHECK11-NEXT: ret void
4758 //
4759 //
4760 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..14
4761 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4762 // CHECK11-NEXT: entry:
4763 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4764 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4765 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4766 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4767 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4768 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4769 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4770 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4771 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4772 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4773 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4774 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4775 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4776 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4777 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4778 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4779 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4780 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4781 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4782 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4783 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4784 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4785 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4786 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4787 // CHECK11: cond.true:
4788 // CHECK11-NEXT: br label [[COND_END:%.*]]
4789 // CHECK11: cond.false:
4790 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4791 // CHECK11-NEXT: br label [[COND_END]]
4792 // CHECK11: cond.end:
4793 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4794 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4795 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4796 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4797 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4798 // CHECK11: omp.inner.for.cond:
4799 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
4800 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !37
4801 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4802 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4803 // CHECK11: omp.inner.for.body:
4804 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !37
4805 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !37
4806 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]]), !llvm.access.group !37
4807 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4808 // CHECK11: omp.inner.for.inc:
4809 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
4810 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !37
4811 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
4812 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !37
4813 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
4814 // CHECK11: omp.inner.for.end:
4815 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4816 // CHECK11: omp.loop.exit:
4817 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4818 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4819 // CHECK11-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
4820 // CHECK11-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4821 // CHECK11: .omp.final.then:
4822 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
4823 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4824 // CHECK11: .omp.final.done:
4825 // CHECK11-NEXT: ret void
4826 //
4827 //
4828 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..15
4829 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4830 // CHECK11-NEXT: entry:
4831 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4832 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4833 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4834 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4835 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4836 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4837 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4838 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4839 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4840 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4841 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4842 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4843 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4844 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4845 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4846 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4847 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4848 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4849 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4850 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4851 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4852 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4853 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
4854 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
4855 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4856 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4857 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4858 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4859 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4860 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4861 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4862 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4863 // CHECK11: cond.true:
4864 // CHECK11-NEXT: br label [[COND_END:%.*]]
4865 // CHECK11: cond.false:
4866 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4867 // CHECK11-NEXT: br label [[COND_END]]
4868 // CHECK11: cond.end:
4869 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4870 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4871 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4872 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4873 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4874 // CHECK11: omp.inner.for.cond:
4875 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
4876 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !40
4877 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4878 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4879 // CHECK11: omp.inner.for.body:
4880 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
4881 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4882 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4883 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !40
4884 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !40
4885 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
4886 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !40
4887 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4888 // CHECK11: omp.body.continue:
4889 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4890 // CHECK11: omp.inner.for.inc:
4891 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
4892 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
4893 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !40
4894 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP41:![0-9]+]]
4895 // CHECK11: omp.inner.for.end:
4896 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4897 // CHECK11: omp.loop.exit:
4898 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4899 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4900 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4901 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4902 // CHECK11: .omp.final.then:
4903 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
4904 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
4905 // CHECK11: .omp.final.done:
4906 // CHECK11-NEXT: ret void
4907 //
4908 //
4909 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
4910 // CHECK11-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4911 // CHECK11-NEXT: entry:
4912 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
4913 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4914 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4915 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4916 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
4917 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4918 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4919 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
4920 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
4921 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4922 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4923 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4924 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
4925 // CHECK11-NEXT: ret void
4926 //
4927 //
4928 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..18
4929 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4930 // CHECK11-NEXT: entry:
4931 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4932 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4933 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4934 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4935 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4936 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4937 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4938 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4939 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4940 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4941 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4942 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4943 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4944 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4945 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4946 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4947 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4948 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4949 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4950 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4951 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4952 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4953 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4954 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4955 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
4956 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4957 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
4958 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4959 // CHECK11: cond.true:
4960 // CHECK11-NEXT: br label [[COND_END:%.*]]
4961 // CHECK11: cond.false:
4962 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4963 // CHECK11-NEXT: br label [[COND_END]]
4964 // CHECK11: cond.end:
4965 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4966 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4967 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4968 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4969 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4970 // CHECK11: omp.inner.for.cond:
4971 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
4972 // CHECK11-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
4973 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4974 // CHECK11: omp.inner.for.body:
4975 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !43
4976 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
4977 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4, !llvm.access.group !43
4978 // CHECK11-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !43
4979 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4, !llvm.access.group !43
4980 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]]), !llvm.access.group !43
4981 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4982 // CHECK11: omp.inner.for.inc:
4983 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
4984 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !43
4985 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4986 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
4987 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !43
4988 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !43
4989 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4990 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !43
4991 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
4992 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !43
4993 // CHECK11-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
4994 // CHECK11-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
4995 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
4996 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
4997 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
4998 // CHECK11: cond.true5:
4999 // CHECK11-NEXT: br label [[COND_END7:%.*]]
5000 // CHECK11: cond.false6:
5001 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
5002 // CHECK11-NEXT: br label [[COND_END7]]
5003 // CHECK11: cond.end7:
5004 // CHECK11-NEXT: [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
5005 // CHECK11-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !43
5006 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !43
5007 // CHECK11-NEXT: store i32 [[TMP20]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
5008 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
5009 // CHECK11: omp.inner.for.end:
5010 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5011 // CHECK11: omp.loop.exit:
5012 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5013 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5014 // CHECK11-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
5015 // CHECK11-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5016 // CHECK11: .omp.final.then:
5017 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
5018 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
5019 // CHECK11: .omp.final.done:
5020 // CHECK11-NEXT: ret void
5021 //
5022 //
5023 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..19
5024 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
5025 // CHECK11-NEXT: entry:
5026 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5027 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5028 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
5029 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
5030 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
5031 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
5032 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5033 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
5034 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5035 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5036 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5037 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5038 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
5039 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5040 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5041 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
5042 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
5043 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
5044 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
5045 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
5046 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5047 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5048 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
5049 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
5050 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
5051 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
5052 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5053 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5054 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5055 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5056 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5057 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5058 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
5059 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5060 // CHECK11: cond.true:
5061 // CHECK11-NEXT: br label [[COND_END:%.*]]
5062 // CHECK11: cond.false:
5063 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5064 // CHECK11-NEXT: br label [[COND_END]]
5065 // CHECK11: cond.end:
5066 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5067 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5068 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5069 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5070 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5071 // CHECK11: omp.inner.for.cond:
5072 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
5073 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !46
5074 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5075 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5076 // CHECK11: omp.inner.for.body:
5077 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
5078 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5079 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5080 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !46
5081 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !46
5082 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
5083 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !46
5084 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5085 // CHECK11: omp.body.continue:
5086 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5087 // CHECK11: omp.inner.for.inc:
5088 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
5089 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
5090 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !46
5091 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP47:![0-9]+]]
5092 // CHECK11: omp.inner.for.end:
5093 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5094 // CHECK11: omp.loop.exit:
5095 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5096 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5097 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5098 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5099 // CHECK11: .omp.final.then:
5100 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
5101 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
5102 // CHECK11: .omp.final.done:
5103 // CHECK11-NEXT: ret void
5104 //
5105 //
5106 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5107 // CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
5108 // CHECK11-NEXT: entry:
5109 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
5110 // CHECK11-NEXT: ret void
5111 //
5112 //
5113 // CHECK13-LABEL: define {{[^@]+}}@main
5114 // CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
5115 // CHECK13-NEXT: entry:
5116 // CHECK13-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5117 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5118 // CHECK13-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
5119 // CHECK13-NEXT: [[N:%.*]] = alloca i32, align 4
5120 // CHECK13-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
5121 // CHECK13-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
5122 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
5123 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
5124 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5125 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5126 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5127 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5128 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
5129 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5130 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
5131 // CHECK13-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
5132 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
5133 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
5134 // CHECK13-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
5135 // CHECK13-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
5136 // CHECK13-NEXT: [[I18:%.*]] = alloca i32, align 4
5137 // CHECK13-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
5138 // CHECK13-NEXT: [[I22:%.*]] = alloca i32, align 4
5139 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
5140 // CHECK13-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
5141 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
5142 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
5143 // CHECK13-NEXT: [[DOTOMP_LB46:%.*]] = alloca i32, align 4
5144 // CHECK13-NEXT: [[DOTOMP_UB47:%.*]] = alloca i32, align 4
5145 // CHECK13-NEXT: [[I48:%.*]] = alloca i32, align 4
5146 // CHECK13-NEXT: [[DOTOMP_IV51:%.*]] = alloca i32, align 4
5147 // CHECK13-NEXT: [[I52:%.*]] = alloca i32, align 4
5148 // CHECK13-NEXT: store i32 0, i32* [[RETVAL]], align 4
5149 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5150 // CHECK13-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
5151 // CHECK13-NEXT: store i32 100, i32* [[N]], align 4
5152 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
5153 // CHECK13-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
5154 // CHECK13-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
5155 // CHECK13-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
5156 // CHECK13-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
5157 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
5158 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
5159 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
5160 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
5161 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5162 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5163 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5164 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5165 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5166 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5167 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5168 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
5169 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
5170 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5171 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
5172 // CHECK13-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
5173 // CHECK13: simd.if.then:
5174 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5175 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5176 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5177 // CHECK13: omp.inner.for.cond:
5178 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5179 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
5180 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5181 // CHECK13-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5182 // CHECK13: omp.inner.for.body:
5183 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5184 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5185 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5186 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
5187 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
5188 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
5189 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
5190 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
5191 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5192 // CHECK13: omp.body.continue:
5193 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5194 // CHECK13: omp.inner.for.inc:
5195 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5196 // CHECK13-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
5197 // CHECK13-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
5198 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
5199 // CHECK13: omp.inner.for.end:
5200 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5201 // CHECK13-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
5202 // CHECK13-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
5203 // CHECK13-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
5204 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
5205 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
5206 // CHECK13-NEXT: br label [[SIMD_IF_END]]
5207 // CHECK13: simd.if.end:
5208 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[N]], align 4
5209 // CHECK13-NEXT: store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
5210 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5211 // CHECK13-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
5212 // CHECK13-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
5213 // CHECK13-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
5214 // CHECK13-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
5215 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
5216 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
5217 // CHECK13-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
5218 // CHECK13-NEXT: store i32 0, i32* [[I18]], align 4
5219 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5220 // CHECK13-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
5221 // CHECK13-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
5222 // CHECK13: simd.if.then20:
5223 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
5224 // CHECK13-NEXT: store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
5225 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
5226 // CHECK13: omp.inner.for.cond23:
5227 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
5228 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
5229 // CHECK13-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
5230 // CHECK13-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
5231 // CHECK13: omp.inner.for.body25:
5232 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
5233 // CHECK13-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
5234 // CHECK13-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
5235 // CHECK13-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
5236 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
5237 // CHECK13-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
5238 // CHECK13-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
5239 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
5240 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
5241 // CHECK13: omp.body.continue30:
5242 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
5243 // CHECK13: omp.inner.for.inc31:
5244 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
5245 // CHECK13-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
5246 // CHECK13-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
5247 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
5248 // CHECK13: omp.inner.for.end33:
5249 // CHECK13-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5250 // CHECK13-NEXT: [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
5251 // CHECK13-NEXT: [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
5252 // CHECK13-NEXT: [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
5253 // CHECK13-NEXT: [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
5254 // CHECK13-NEXT: store i32 [[ADD37]], i32* [[I22]], align 4
5255 // CHECK13-NEXT: br label [[SIMD_IF_END38]]
5256 // CHECK13: simd.if.end38:
5257 // CHECK13-NEXT: [[TMP25:%.*]] = load i32, i32* [[M]], align 4
5258 // CHECK13-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
5259 // CHECK13-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
5260 // CHECK13-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
5261 // CHECK13-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
5262 // CHECK13-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
5263 // CHECK13-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
5264 // CHECK13-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
5265 // CHECK13-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
5266 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB46]], align 4
5267 // CHECK13-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
5268 // CHECK13-NEXT: store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
5269 // CHECK13-NEXT: store i32 0, i32* [[I48]], align 4
5270 // CHECK13-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
5271 // CHECK13-NEXT: [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
5272 // CHECK13-NEXT: br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
5273 // CHECK13: simd.if.then50:
5274 // CHECK13-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
5275 // CHECK13-NEXT: store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
5276 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND53:%.*]]
5277 // CHECK13: omp.inner.for.cond53:
5278 // CHECK13-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
5279 // CHECK13-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
5280 // CHECK13-NEXT: [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
5281 // CHECK13-NEXT: br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
5282 // CHECK13: omp.inner.for.body55:
5283 // CHECK13-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
5284 // CHECK13-NEXT: [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
5285 // CHECK13-NEXT: [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
5286 // CHECK13-NEXT: store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
5287 // CHECK13-NEXT: [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
5288 // CHECK13-NEXT: [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
5289 // CHECK13-NEXT: [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
5290 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
5291 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE60:%.*]]
5292 // CHECK13: omp.body.continue60:
5293 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC61:%.*]]
5294 // CHECK13: omp.inner.for.inc61:
5295 // CHECK13-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
5296 // CHECK13-NEXT: [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
5297 // CHECK13-NEXT: store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
5298 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
5299 // CHECK13: omp.inner.for.end63:
5300 // CHECK13-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
5301 // CHECK13-NEXT: [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
5302 // CHECK13-NEXT: [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
5303 // CHECK13-NEXT: [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
5304 // CHECK13-NEXT: [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
5305 // CHECK13-NEXT: store i32 [[ADD67]], i32* [[I52]], align 4
5306 // CHECK13-NEXT: br label [[SIMD_IF_END68]]
5307 // CHECK13: simd.if.end68:
5308 // CHECK13-NEXT: [[TMP37:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5309 // CHECK13-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP37]])
5310 // CHECK13-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
5311 // CHECK13-NEXT: [[TMP38:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
5312 // CHECK13-NEXT: call void @llvm.stackrestore(i8* [[TMP38]])
5313 // CHECK13-NEXT: [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
5314 // CHECK13-NEXT: ret i32 [[TMP39]]
5315 //
5316 //
5317 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5318 // CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
5319 // CHECK13-NEXT: entry:
5320 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5321 // CHECK13-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5322 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
5323 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
5324 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5325 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5326 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5327 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
5328 // CHECK13-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
5329 // CHECK13-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
5330 // CHECK13-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
5331 // CHECK13-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
5332 // CHECK13-NEXT: [[I6:%.*]] = alloca i32, align 4
5333 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5334 // CHECK13-NEXT: [[_TMP18:%.*]] = alloca i32, align 4
5335 // CHECK13-NEXT: [[DOTOMP_LB19:%.*]] = alloca i32, align 4
5336 // CHECK13-NEXT: [[DOTOMP_UB20:%.*]] = alloca i32, align 4
5337 // CHECK13-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
5338 // CHECK13-NEXT: [[I22:%.*]] = alloca i32, align 4
5339 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5340 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
5341 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5342 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5343 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5344 // CHECK13-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
5345 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5346 // CHECK13: omp.inner.for.cond:
5347 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
5348 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
5349 // CHECK13-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
5350 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5351 // CHECK13: omp.inner.for.body:
5352 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
5353 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
5354 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5355 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
5356 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
5357 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
5358 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5359 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
5360 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5361 // CHECK13: omp.body.continue:
5362 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5363 // CHECK13: omp.inner.for.inc:
5364 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
5365 // CHECK13-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5366 // CHECK13-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
5367 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
5368 // CHECK13: omp.inner.for.end:
5369 // CHECK13-NEXT: store i32 10, i32* [[I]], align 4
5370 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
5371 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
5372 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
5373 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
5374 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
5375 // CHECK13: omp.inner.for.cond7:
5376 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
5377 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !15
5378 // CHECK13-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5379 // CHECK13-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
5380 // CHECK13: omp.inner.for.body9:
5381 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
5382 // CHECK13-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
5383 // CHECK13-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
5384 // CHECK13-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !15
5385 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !15
5386 // CHECK13-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
5387 // CHECK13-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
5388 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !15
5389 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
5390 // CHECK13: omp.body.continue14:
5391 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
5392 // CHECK13: omp.inner.for.inc15:
5393 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
5394 // CHECK13-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
5395 // CHECK13-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
5396 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
5397 // CHECK13: omp.inner.for.end17:
5398 // CHECK13-NEXT: store i32 10, i32* [[I6]], align 4
5399 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
5400 // CHECK13-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
5401 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB19]], align 4
5402 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB20]], align 4
5403 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
5404 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
5405 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
5406 // CHECK13: omp.inner.for.cond23:
5407 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
5408 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !18
5409 // CHECK13-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5410 // CHECK13-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
5411 // CHECK13: omp.inner.for.body25:
5412 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
5413 // CHECK13-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
5414 // CHECK13-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
5415 // CHECK13-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !18
5416 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !18
5417 // CHECK13-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
5418 // CHECK13-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
5419 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !18
5420 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
5421 // CHECK13: omp.body.continue30:
5422 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
5423 // CHECK13: omp.inner.for.inc31:
5424 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
5425 // CHECK13-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
5426 // CHECK13-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
5427 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP19:![0-9]+]]
5428 // CHECK13: omp.inner.for.end33:
5429 // CHECK13-NEXT: store i32 10, i32* [[I22]], align 4
5430 // CHECK13-NEXT: ret i32 0
5431 //
5432 //
5433 // CHECK15-LABEL: define {{[^@]+}}@main
5434 // CHECK15-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
5435 // CHECK15-NEXT: entry:
5436 // CHECK15-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5437 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5438 // CHECK15-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
5439 // CHECK15-NEXT: [[N:%.*]] = alloca i32, align 4
5440 // CHECK15-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
5441 // CHECK15-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
5442 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
5443 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
5444 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5445 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5446 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5447 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5448 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
5449 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5450 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
5451 // CHECK15-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
5452 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
5453 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
5454 // CHECK15-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
5455 // CHECK15-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
5456 // CHECK15-NEXT: [[I18:%.*]] = alloca i32, align 4
5457 // CHECK15-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
5458 // CHECK15-NEXT: [[I22:%.*]] = alloca i32, align 4
5459 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
5460 // CHECK15-NEXT: [[_TMP39:%.*]] = alloca i32, align 4
5461 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
5462 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
5463 // CHECK15-NEXT: [[DOTOMP_LB45:%.*]] = alloca i32, align 4
5464 // CHECK15-NEXT: [[DOTOMP_UB46:%.*]] = alloca i32, align 4
5465 // CHECK15-NEXT: [[I47:%.*]] = alloca i32, align 4
5466 // CHECK15-NEXT: [[DOTOMP_IV50:%.*]] = alloca i32, align 4
5467 // CHECK15-NEXT: [[I51:%.*]] = alloca i32, align 4
5468 // CHECK15-NEXT: store i32 0, i32* [[RETVAL]], align 4
5469 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5470 // CHECK15-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
5471 // CHECK15-NEXT: store i32 100, i32* [[N]], align 4
5472 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
5473 // CHECK15-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
5474 // CHECK15-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
5475 // CHECK15-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
5476 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
5477 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
5478 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
5479 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
5480 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5481 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
5482 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5483 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5484 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5485 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5486 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5487 // CHECK15-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
5488 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
5489 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5490 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5491 // CHECK15-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
5492 // CHECK15: simd.if.then:
5493 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5494 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5495 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5496 // CHECK15: omp.inner.for.cond:
5497 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5498 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
5499 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5500 // CHECK15-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5501 // CHECK15: omp.inner.for.body:
5502 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5503 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5504 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5505 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
5506 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
5507 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
5508 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
5509 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5510 // CHECK15: omp.body.continue:
5511 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5512 // CHECK15: omp.inner.for.inc:
5513 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5514 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
5515 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
5516 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
5517 // CHECK15: omp.inner.for.end:
5518 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5519 // CHECK15-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
5520 // CHECK15-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
5521 // CHECK15-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
5522 // CHECK15-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
5523 // CHECK15-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
5524 // CHECK15-NEXT: br label [[SIMD_IF_END]]
5525 // CHECK15: simd.if.end:
5526 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[N]], align 4
5527 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
5528 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5529 // CHECK15-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
5530 // CHECK15-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
5531 // CHECK15-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
5532 // CHECK15-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
5533 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
5534 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
5535 // CHECK15-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
5536 // CHECK15-NEXT: store i32 0, i32* [[I18]], align 4
5537 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5538 // CHECK15-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
5539 // CHECK15-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
5540 // CHECK15: simd.if.then20:
5541 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
5542 // CHECK15-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
5543 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
5544 // CHECK15: omp.inner.for.cond23:
5545 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
5546 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
5547 // CHECK15-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
5548 // CHECK15-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
5549 // CHECK15: omp.inner.for.body25:
5550 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
5551 // CHECK15-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
5552 // CHECK15-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
5553 // CHECK15-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
5554 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
5555 // CHECK15-NEXT: [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
5556 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
5557 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE29:%.*]]
5558 // CHECK15: omp.body.continue29:
5559 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC30:%.*]]
5560 // CHECK15: omp.inner.for.inc30:
5561 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
5562 // CHECK15-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
5563 // CHECK15-NEXT: store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
5564 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
5565 // CHECK15: omp.inner.for.end32:
5566 // CHECK15-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
5567 // CHECK15-NEXT: [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
5568 // CHECK15-NEXT: [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
5569 // CHECK15-NEXT: [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
5570 // CHECK15-NEXT: [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
5571 // CHECK15-NEXT: store i32 [[ADD36]], i32* [[I22]], align 4
5572 // CHECK15-NEXT: br label [[SIMD_IF_END37]]
5573 // CHECK15: simd.if.end37:
5574 // CHECK15-NEXT: [[TMP24:%.*]] = load i32, i32* [[M]], align 4
5575 // CHECK15-NEXT: store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
5576 // CHECK15-NEXT: [[TMP25:%.*]] = load i32, i32* [[N]], align 4
5577 // CHECK15-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
5578 // CHECK15-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
5579 // CHECK15-NEXT: [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
5580 // CHECK15-NEXT: [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
5581 // CHECK15-NEXT: [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
5582 // CHECK15-NEXT: store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
5583 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB45]], align 4
5584 // CHECK15-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
5585 // CHECK15-NEXT: store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
5586 // CHECK15-NEXT: store i32 0, i32* [[I47]], align 4
5587 // CHECK15-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
5588 // CHECK15-NEXT: [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
5589 // CHECK15-NEXT: br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
5590 // CHECK15: simd.if.then49:
5591 // CHECK15-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
5592 // CHECK15-NEXT: store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
5593 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND52:%.*]]
5594 // CHECK15: omp.inner.for.cond52:
5595 // CHECK15-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
5596 // CHECK15-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
5597 // CHECK15-NEXT: [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
5598 // CHECK15-NEXT: br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
5599 // CHECK15: omp.inner.for.body54:
5600 // CHECK15-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
5601 // CHECK15-NEXT: [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
5602 // CHECK15-NEXT: [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
5603 // CHECK15-NEXT: store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
5604 // CHECK15-NEXT: [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
5605 // CHECK15-NEXT: [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
5606 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
5607 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE58:%.*]]
5608 // CHECK15: omp.body.continue58:
5609 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC59:%.*]]
5610 // CHECK15: omp.inner.for.inc59:
5611 // CHECK15-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
5612 // CHECK15-NEXT: [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
5613 // CHECK15-NEXT: store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
5614 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
5615 // CHECK15: omp.inner.for.end61:
5616 // CHECK15-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
5617 // CHECK15-NEXT: [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
5618 // CHECK15-NEXT: [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
5619 // CHECK15-NEXT: [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
5620 // CHECK15-NEXT: [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
5621 // CHECK15-NEXT: store i32 [[ADD65]], i32* [[I51]], align 4
5622 // CHECK15-NEXT: br label [[SIMD_IF_END66]]
5623 // CHECK15: simd.if.end66:
5624 // CHECK15-NEXT: [[TMP36:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
5625 // CHECK15-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP36]])
5626 // CHECK15-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
5627 // CHECK15-NEXT: [[TMP37:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
5628 // CHECK15-NEXT: call void @llvm.stackrestore(i8* [[TMP37]])
5629 // CHECK15-NEXT: [[TMP38:%.*]] = load i32, i32* [[RETVAL]], align 4
5630 // CHECK15-NEXT: ret i32 [[TMP38]]
5631 //
5632 //
5633 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5634 // CHECK15-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
5635 // CHECK15-NEXT: entry:
5636 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5637 // CHECK15-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5638 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
5639 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
5640 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5641 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5642 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5643 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
5644 // CHECK15-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
5645 // CHECK15-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
5646 // CHECK15-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
5647 // CHECK15-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
5648 // CHECK15-NEXT: [[I6:%.*]] = alloca i32, align 4
5649 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5650 // CHECK15-NEXT: [[_TMP17:%.*]] = alloca i32, align 4
5651 // CHECK15-NEXT: [[DOTOMP_LB18:%.*]] = alloca i32, align 4
5652 // CHECK15-NEXT: [[DOTOMP_UB19:%.*]] = alloca i32, align 4
5653 // CHECK15-NEXT: [[DOTOMP_IV20:%.*]] = alloca i32, align 4
5654 // CHECK15-NEXT: [[I21:%.*]] = alloca i32, align 4
5655 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5656 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
5657 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5658 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5659 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5660 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
5661 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5662 // CHECK15: omp.inner.for.cond:
5663 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5664 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
5665 // CHECK15-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
5666 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5667 // CHECK15: omp.inner.for.body:
5668 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5669 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
5670 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5671 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
5672 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
5673 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
5674 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
5675 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5676 // CHECK15: omp.body.continue:
5677 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5678 // CHECK15: omp.inner.for.inc:
5679 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5680 // CHECK15-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
5681 // CHECK15-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5682 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
5683 // CHECK15: omp.inner.for.end:
5684 // CHECK15-NEXT: store i32 10, i32* [[I]], align 4
5685 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
5686 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
5687 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
5688 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
5689 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
5690 // CHECK15: omp.inner.for.cond7:
5691 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
5692 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
5693 // CHECK15-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5694 // CHECK15-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
5695 // CHECK15: omp.inner.for.body9:
5696 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
5697 // CHECK15-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
5698 // CHECK15-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
5699 // CHECK15-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
5700 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !16
5701 // CHECK15-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
5702 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !16
5703 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE13:%.*]]
5704 // CHECK15: omp.body.continue13:
5705 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC14:%.*]]
5706 // CHECK15: omp.inner.for.inc14:
5707 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
5708 // CHECK15-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
5709 // CHECK15-NEXT: store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
5710 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
5711 // CHECK15: omp.inner.for.end16:
5712 // CHECK15-NEXT: store i32 10, i32* [[I6]], align 4
5713 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
5714 // CHECK15-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
5715 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB18]], align 4
5716 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB19]], align 4
5717 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
5718 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
5719 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND22:%.*]]
5720 // CHECK15: omp.inner.for.cond22:
5721 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
5722 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !19
5723 // CHECK15-NEXT: [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5724 // CHECK15-NEXT: br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
5725 // CHECK15: omp.inner.for.body24:
5726 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
5727 // CHECK15-NEXT: [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
5728 // CHECK15-NEXT: [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
5729 // CHECK15-NEXT: store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !19
5730 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !19
5731 // CHECK15-NEXT: [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
5732 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !19
5733 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE28:%.*]]
5734 // CHECK15: omp.body.continue28:
5735 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC29:%.*]]
5736 // CHECK15: omp.inner.for.inc29:
5737 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
5738 // CHECK15-NEXT: [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
5739 // CHECK15-NEXT: store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
5740 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP20:![0-9]+]]
5741 // CHECK15: omp.inner.for.end31:
5742 // CHECK15-NEXT: store i32 10, i32* [[I21]], align 4
5743 // CHECK15-NEXT: ret i32 0
5744 //
5745