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][Y];
25
fooSS26 int foo(void) {
27
28 #pragma omp target teams distribute simd collapse(2)
29 for(int i = 0; i < X; i++) {
30 for(int j = 0; j < Y; j++) {
31 a[i][j] = (T)0;
32 }
33 }
34
35 // discard loop variables not needed here
36
37 return a[0][0];
38 }
39 };
40
teams_template_struct(void)41 int teams_template_struct(void) {
42 SS<int, 123, 456> V;
43 return V.foo();
44
45 }
46 #endif // CK1
47
48 // Test host codegen.
49 // 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
50 // 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
51 // 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
52 // 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
53 // 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
54 // 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
55
56 // 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
57 // 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
58 // 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
59 // 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
60 // 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
61 // 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
62 #ifdef CK2
63
64 template <typename T, int n, int m>
tmain(T argc)65 int tmain(T argc) {
66 T a[n][m];
67 #pragma omp target teams distribute simd collapse(2)
68 for(int i = 0; i < n; i++) {
69 for(int j = 0; j < m; j++) {
70 a[i][j] = (T)0;
71 }
72 }
73 return 0;
74 }
75
main(int argc,char ** argv)76 int main (int argc, char **argv) {
77 int n = 100;
78 int m = 2;
79 int a[n][m];
80 #pragma omp target teams distribute simd collapse(2)
81 for(int i = 0; i < n; i++) {
82 for(int j = 0; j < m; j++) {
83 a[i][j] = 0;
84 }
85 }
86 return tmain<int, 10, 2>(argc);
87 }
88
89
90
91
92
93 // discard loop variables not needed here
94
95 #endif // CK2
96 #endif // #ifndef HEADER
97 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
98 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
99 // CHECK1-NEXT: entry:
100 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
101 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[V]])
102 // CHECK1-NEXT: ret i32 [[CALL]]
103 //
104 //
105 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
106 // CHECK1-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
107 // CHECK1-NEXT: entry:
108 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
109 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
110 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
111 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
112 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
113 // CHECK1-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
114 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
115 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
116 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
117 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
118 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
119 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
120 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
121 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x [456 x i32]]**
122 // CHECK1-NEXT: store [123 x [456 x i32]]* [[A]], [123 x [456 x i32]]** [[TMP3]], align 8
123 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
124 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
125 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
126 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
127 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
128 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
129 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
130 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
131 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
132 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
133 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
134 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
135 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
136 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
137 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
138 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
139 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
140 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
141 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
142 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
143 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
144 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
145 // CHECK1-NEXT: store i64 56088, i64* [[TMP15]], align 8
146 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
147 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
148 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
149 // CHECK1: omp_offload.failed:
150 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR3:[0-9]+]]
151 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
152 // CHECK1: omp_offload.cont:
153 // CHECK1-NEXT: [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
154 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A3]], i64 0, i64 0
155 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i64 0, i64 0
156 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
157 // CHECK1-NEXT: ret i32 [[TMP18]]
158 //
159 //
160 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
161 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
162 // CHECK1-NEXT: entry:
163 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
164 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
165 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
166 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
167 // CHECK1-NEXT: ret void
168 //
169 //
170 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
171 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
172 // CHECK1-NEXT: entry:
173 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
174 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
175 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
176 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
177 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
178 // CHECK1-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
179 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
180 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
181 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
182 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
183 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
184 // CHECK1-NEXT: [[J:%.*]] = alloca i32, align 4
185 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
186 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
187 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
188 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
189 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
190 // CHECK1-NEXT: store i32 56087, i32* [[DOTOMP_UB]], align 4
191 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
192 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
193 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
194 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
195 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
196 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
197 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
198 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
199 // CHECK1: cond.true:
200 // CHECK1-NEXT: br label [[COND_END:%.*]]
201 // CHECK1: cond.false:
202 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
203 // CHECK1-NEXT: br label [[COND_END]]
204 // CHECK1: cond.end:
205 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
206 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
207 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
208 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
209 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
210 // CHECK1: omp.inner.for.cond:
211 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
212 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !4
213 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
214 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
215 // CHECK1: omp.inner.for.body:
216 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
217 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
218 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
219 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
220 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !4
221 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
222 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
223 // CHECK1-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
224 // CHECK1-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
225 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
226 // CHECK1-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
227 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
228 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !4
229 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
230 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !4
231 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
232 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i64 0, i64 [[IDXPROM]]
233 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !4
234 // CHECK1-NEXT: [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
235 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
236 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX8]], align 4, !llvm.access.group !4
237 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
238 // CHECK1: omp.body.continue:
239 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
240 // CHECK1: omp.inner.for.inc:
241 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
242 // CHECK1-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
243 // CHECK1-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !4
244 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
245 // CHECK1: omp.inner.for.end:
246 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
247 // CHECK1: omp.loop.exit:
248 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
249 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
250 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
251 // CHECK1-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
252 // CHECK1: .omp.final.then:
253 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
254 // CHECK1-NEXT: store i32 456, i32* [[J]], align 4
255 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
256 // CHECK1: .omp.final.done:
257 // CHECK1-NEXT: ret void
258 //
259 //
260 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
261 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
262 // CHECK1-NEXT: entry:
263 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
264 // CHECK1-NEXT: ret void
265 //
266 //
267 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
268 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
269 // CHECK3-NEXT: entry:
270 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
271 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[V]])
272 // CHECK3-NEXT: ret i32 [[CALL]]
273 //
274 //
275 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
276 // CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
277 // CHECK3-NEXT: entry:
278 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
279 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
280 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
281 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
282 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
283 // CHECK3-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
284 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
285 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
286 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
287 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
288 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
289 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
290 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
291 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x [456 x i32]]**
292 // CHECK3-NEXT: store [123 x [456 x i32]]* [[A]], [123 x [456 x i32]]** [[TMP3]], align 4
293 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
294 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
295 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
296 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
297 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
298 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
299 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
300 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
301 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
302 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
303 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
304 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
305 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
306 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
307 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
308 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
309 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
310 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
311 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
312 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
313 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
314 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
315 // CHECK3-NEXT: store i64 56088, i64* [[TMP15]], align 8
316 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
317 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
318 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
319 // CHECK3: omp_offload.failed:
320 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR3:[0-9]+]]
321 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
322 // CHECK3: omp_offload.cont:
323 // CHECK3-NEXT: [[A3:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
324 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A3]], i32 0, i32 0
325 // CHECK3-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i32 0, i32 0
326 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX4]], align 4
327 // CHECK3-NEXT: ret i32 [[TMP18]]
328 //
329 //
330 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
331 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
332 // CHECK3-NEXT: entry:
333 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
334 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
335 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
336 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
337 // CHECK3-NEXT: ret void
338 //
339 //
340 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
341 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR2:[0-9]+]] {
342 // CHECK3-NEXT: entry:
343 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
344 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
345 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
346 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
347 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
348 // CHECK3-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
349 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
350 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
351 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
352 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
353 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
354 // CHECK3-NEXT: [[J:%.*]] = alloca i32, align 4
355 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
356 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
357 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
358 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
359 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
360 // CHECK3-NEXT: store i32 56087, i32* [[DOTOMP_UB]], align 4
361 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
362 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
363 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
364 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
365 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
366 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
367 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 56087
368 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
369 // CHECK3: cond.true:
370 // CHECK3-NEXT: br label [[COND_END:%.*]]
371 // CHECK3: cond.false:
372 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
373 // CHECK3-NEXT: br label [[COND_END]]
374 // CHECK3: cond.end:
375 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 56087, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
376 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
377 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
378 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
379 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
380 // CHECK3: omp.inner.for.cond:
381 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
382 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !5
383 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
384 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
385 // CHECK3: omp.inner.for.body:
386 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
387 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP8]], 456
388 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
389 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
390 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !5
391 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
392 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
393 // CHECK3-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP10]], 456
394 // CHECK3-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
395 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
396 // CHECK3-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
397 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
398 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !5
399 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
400 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !5
401 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i32 0, i32 [[TMP11]]
402 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !5
403 // CHECK3-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP12]]
404 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX7]], align 4, !llvm.access.group !5
405 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
406 // CHECK3: omp.body.continue:
407 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
408 // CHECK3: omp.inner.for.inc:
409 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
410 // CHECK3-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
411 // CHECK3-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !5
412 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
413 // CHECK3: omp.inner.for.end:
414 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
415 // CHECK3: omp.loop.exit:
416 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
417 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
418 // CHECK3-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
419 // CHECK3-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
420 // CHECK3: .omp.final.then:
421 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
422 // CHECK3-NEXT: store i32 456, i32* [[J]], align 4
423 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
424 // CHECK3: .omp.final.done:
425 // CHECK3-NEXT: ret void
426 //
427 //
428 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
429 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] {
430 // CHECK3-NEXT: entry:
431 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
432 // CHECK3-NEXT: ret void
433 //
434 //
435 // CHECK5-LABEL: define {{[^@]+}}@_Z21teams_template_structv
436 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
437 // CHECK5-NEXT: entry:
438 // CHECK5-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
439 // CHECK5-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[V]])
440 // CHECK5-NEXT: ret i32 [[CALL]]
441 //
442 //
443 // CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
444 // CHECK5-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
445 // CHECK5-NEXT: entry:
446 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
447 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
448 // CHECK5-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
449 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
450 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
451 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
452 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
453 // CHECK5-NEXT: [[J:%.*]] = alloca i32, align 4
454 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
455 // CHECK5-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
456 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
457 // CHECK5-NEXT: store i32 56087, i32* [[DOTOMP_UB]], align 4
458 // CHECK5-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
459 // CHECK5-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
460 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
461 // CHECK5: omp.inner.for.cond:
462 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
463 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
464 // CHECK5-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
465 // CHECK5-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
466 // CHECK5: omp.inner.for.body:
467 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
468 // CHECK5-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP3]], 456
469 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
470 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
471 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
472 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
473 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
474 // CHECK5-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP5]], 456
475 // CHECK5-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
476 // CHECK5-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], [[MUL4]]
477 // CHECK5-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
478 // CHECK5-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
479 // CHECK5-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !2
480 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
481 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
482 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
483 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i64 0, i64 [[IDXPROM]]
484 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !2
485 // CHECK5-NEXT: [[IDXPROM7:%.*]] = sext i32 [[TMP7]] to i64
486 // CHECK5-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
487 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX8]], align 4, !llvm.access.group !2
488 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
489 // CHECK5: omp.body.continue:
490 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
491 // CHECK5: omp.inner.for.inc:
492 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
493 // CHECK5-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP8]], 1
494 // CHECK5-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
495 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
496 // CHECK5: omp.inner.for.end:
497 // CHECK5-NEXT: store i32 123, i32* [[I]], align 4
498 // CHECK5-NEXT: store i32 456, i32* [[J]], align 4
499 // CHECK5-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
500 // CHECK5-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A10]], i64 0, i64 0
501 // CHECK5-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX11]], i64 0, i64 0
502 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX12]], align 4
503 // CHECK5-NEXT: ret i32 [[TMP9]]
504 //
505 //
506 // CHECK7-LABEL: define {{[^@]+}}@_Z21teams_template_structv
507 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
508 // CHECK7-NEXT: entry:
509 // CHECK7-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
510 // CHECK7-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[V]])
511 // CHECK7-NEXT: ret i32 [[CALL]]
512 //
513 //
514 // CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
515 // CHECK7-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(224352) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
516 // CHECK7-NEXT: entry:
517 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
518 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
519 // CHECK7-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
520 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
521 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
522 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
523 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
524 // CHECK7-NEXT: [[J:%.*]] = alloca i32, align 4
525 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
526 // CHECK7-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
527 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
528 // CHECK7-NEXT: store i32 56087, i32* [[DOTOMP_UB]], align 4
529 // CHECK7-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
530 // CHECK7-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
531 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
532 // CHECK7: omp.inner.for.cond:
533 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
534 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
535 // CHECK7-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
536 // CHECK7-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
537 // CHECK7: omp.inner.for.body:
538 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
539 // CHECK7-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP3]], 456
540 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
541 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
542 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
543 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
544 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
545 // CHECK7-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP5]], 456
546 // CHECK7-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 456
547 // CHECK7-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], [[MUL4]]
548 // CHECK7-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
549 // CHECK7-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
550 // CHECK7-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !3
551 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
552 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
553 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A]], i32 0, i32 [[TMP6]]
554 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !3
555 // CHECK7-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP7]]
556 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX7]], align 4, !llvm.access.group !3
557 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
558 // CHECK7: omp.body.continue:
559 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
560 // CHECK7: omp.inner.for.inc:
561 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
562 // CHECK7-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP8]], 1
563 // CHECK7-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
564 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
565 // CHECK7: omp.inner.for.end:
566 // CHECK7-NEXT: store i32 123, i32* [[I]], align 4
567 // CHECK7-NEXT: store i32 456, i32* [[J]], align 4
568 // CHECK7-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
569 // CHECK7-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds [123 x [456 x i32]], [123 x [456 x i32]]* [[A9]], i32 0, i32 0
570 // CHECK7-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds [456 x i32], [456 x i32]* [[ARRAYIDX10]], i32 0, i32 0
571 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX11]], align 4
572 // CHECK7-NEXT: ret i32 [[TMP9]]
573 //
574 //
575 // CHECK9-LABEL: define {{[^@]+}}@main
576 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
577 // CHECK9-NEXT: entry:
578 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
579 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
580 // CHECK9-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
581 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
582 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
583 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
584 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
585 // CHECK9-NEXT: [[__VLA_EXPR1:%.*]] = alloca i64, align 8
586 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
587 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
588 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 8
589 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 8
590 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 8
591 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 8
592 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
593 // CHECK9-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
594 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
595 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
596 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_4:%.*]] = alloca i64, align 8
597 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
598 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
599 // CHECK9-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
600 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
601 // CHECK9-NEXT: store i32 2, i32* [[M]], align 4
602 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
603 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
604 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[M]], align 4
605 // CHECK9-NEXT: [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
606 // CHECK9-NEXT: [[TMP4:%.*]] = call i8* @llvm.stacksave()
607 // CHECK9-NEXT: store i8* [[TMP4]], i8** [[SAVED_STACK]], align 8
608 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
609 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
610 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
611 // CHECK9-NEXT: store i64 [[TMP3]], i64* [[__VLA_EXPR1]], align 8
612 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[N]], align 4
613 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
614 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[CONV]], align 4
615 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[N_CASTED]], align 8
616 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[M]], align 4
617 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[M_CASTED]] to i32*
618 // CHECK9-NEXT: store i32 [[TMP8]], i32* [[CONV1]], align 4
619 // CHECK9-NEXT: [[TMP9:%.*]] = load i64, i64* [[M_CASTED]], align 8
620 // CHECK9-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
621 // CHECK9-NEXT: [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
622 // CHECK9-NEXT: [[TMP12:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
623 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP12]], i8* align 8 bitcast ([5 x i64]* @.offload_sizes to i8*), i64 40, i1 false)
624 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
625 // CHECK9-NEXT: [[TMP14:%.*]] = bitcast i8** [[TMP13]] to i64*
626 // CHECK9-NEXT: store i64 [[TMP7]], i64* [[TMP14]], align 8
627 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
628 // CHECK9-NEXT: [[TMP16:%.*]] = bitcast i8** [[TMP15]] to i64*
629 // CHECK9-NEXT: store i64 [[TMP7]], i64* [[TMP16]], align 8
630 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
631 // CHECK9-NEXT: store i8* null, i8** [[TMP17]], align 8
632 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
633 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i64*
634 // CHECK9-NEXT: store i64 [[TMP9]], i64* [[TMP19]], align 8
635 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
636 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
637 // CHECK9-NEXT: store i64 [[TMP9]], i64* [[TMP21]], align 8
638 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
639 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
640 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
641 // CHECK9-NEXT: [[TMP24:%.*]] = bitcast i8** [[TMP23]] to i64*
642 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP24]], align 8
643 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
644 // CHECK9-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to i64*
645 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP26]], align 8
646 // CHECK9-NEXT: [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
647 // CHECK9-NEXT: store i8* null, i8** [[TMP27]], align 8
648 // CHECK9-NEXT: [[TMP28:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
649 // CHECK9-NEXT: [[TMP29:%.*]] = bitcast i8** [[TMP28]] to i64*
650 // CHECK9-NEXT: store i64 [[TMP3]], i64* [[TMP29]], align 8
651 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
652 // CHECK9-NEXT: [[TMP31:%.*]] = bitcast i8** [[TMP30]] to i64*
653 // CHECK9-NEXT: store i64 [[TMP3]], i64* [[TMP31]], align 8
654 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
655 // CHECK9-NEXT: store i8* null, i8** [[TMP32]], align 8
656 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
657 // CHECK9-NEXT: [[TMP34:%.*]] = bitcast i8** [[TMP33]] to i32**
658 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP34]], align 8
659 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
660 // CHECK9-NEXT: [[TMP36:%.*]] = bitcast i8** [[TMP35]] to i32**
661 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP36]], align 8
662 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
663 // CHECK9-NEXT: store i64 [[TMP11]], i64* [[TMP37]], align 8
664 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
665 // CHECK9-NEXT: store i8* null, i8** [[TMP38]], align 8
666 // CHECK9-NEXT: [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
667 // CHECK9-NEXT: [[TMP40:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
668 // CHECK9-NEXT: [[TMP41:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
669 // CHECK9-NEXT: [[TMP42:%.*]] = load i32, i32* [[N]], align 4
670 // CHECK9-NEXT: store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_]], align 4
671 // CHECK9-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
672 // CHECK9-NEXT: store i32 [[TMP43]], i32* [[DOTCAPTURE_EXPR_3]], align 4
673 // CHECK9-NEXT: [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
674 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP44]], 0
675 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
676 // CHECK9-NEXT: [[CONV5:%.*]] = sext i32 [[DIV]] to i64
677 // CHECK9-NEXT: [[TMP45:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
678 // CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP45]], 0
679 // CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
680 // CHECK9-NEXT: [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
681 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV5]], [[CONV8]]
682 // CHECK9-NEXT: [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
683 // CHECK9-NEXT: store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_4]], align 8
684 // CHECK9-NEXT: [[TMP46:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_4]], align 8
685 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i64 [[TMP46]], 1
686 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
687 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
688 // CHECK9-NEXT: store i32 1, i32* [[TMP47]], align 4
689 // CHECK9-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
690 // CHECK9-NEXT: store i32 5, i32* [[TMP48]], align 4
691 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
692 // CHECK9-NEXT: store i8** [[TMP39]], i8*** [[TMP49]], align 8
693 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
694 // CHECK9-NEXT: store i8** [[TMP40]], i8*** [[TMP50]], align 8
695 // CHECK9-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
696 // CHECK9-NEXT: store i64* [[TMP41]], i64** [[TMP51]], align 8
697 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
698 // CHECK9-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP52]], align 8
699 // CHECK9-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
700 // CHECK9-NEXT: store i8** null, i8*** [[TMP53]], align 8
701 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
702 // CHECK9-NEXT: store i8** null, i8*** [[TMP54]], align 8
703 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
704 // CHECK9-NEXT: store i64 [[ADD]], i64* [[TMP55]], align 8
705 // CHECK9-NEXT: [[TMP56:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
706 // CHECK9-NEXT: [[TMP57:%.*]] = icmp ne i32 [[TMP56]], 0
707 // CHECK9-NEXT: br i1 [[TMP57]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
708 // CHECK9: omp_offload.failed:
709 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80(i64 [[TMP7]], i64 [[TMP9]], i64 [[TMP1]], i64 [[TMP3]], i32* [[VLA]]) #[[ATTR4:[0-9]+]]
710 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
711 // CHECK9: omp_offload.cont:
712 // CHECK9-NEXT: [[TMP58:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
713 // CHECK9-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP58]])
714 // CHECK9-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
715 // CHECK9-NEXT: [[TMP59:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
716 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP59]])
717 // CHECK9-NEXT: [[TMP60:%.*]] = load i32, i32* [[RETVAL]], align 4
718 // CHECK9-NEXT: ret i32 [[TMP60]]
719 //
720 //
721 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80
722 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
723 // CHECK9-NEXT: entry:
724 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
725 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
726 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
727 // CHECK9-NEXT: [[VLA_ADDR2:%.*]] = alloca i64, align 8
728 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
729 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
730 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
731 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
732 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
733 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
734 // CHECK9-NEXT: store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
735 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
736 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
737 // CHECK9-NEXT: [[CONV3:%.*]] = bitcast i64* [[M_ADDR]] to i32*
738 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
739 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
740 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
741 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
742 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED]] to i32*
743 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV4]], align 4
744 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
745 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[CONV3]], align 4
746 // CHECK9-NEXT: [[CONV5:%.*]] = bitcast i64* [[M_CASTED]] to i32*
747 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[CONV5]], align 4
748 // CHECK9-NEXT: [[TMP6:%.*]] = load i64, i64* [[M_CASTED]], align 8
749 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i64, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i64 [[TMP4]], i64 [[TMP6]], i64 [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
750 // CHECK9-NEXT: ret void
751 //
752 //
753 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
754 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], i64 noundef [[M:%.*]], i64 noundef [[VLA:%.*]], i64 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
755 // CHECK9-NEXT: entry:
756 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
757 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
758 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
759 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
760 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
761 // CHECK9-NEXT: [[VLA_ADDR2:%.*]] = alloca i64, align 8
762 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
763 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
764 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
765 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
766 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
767 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32, align 4
768 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_6:%.*]] = alloca i64, align 8
769 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
770 // CHECK9-NEXT: [[J:%.*]] = alloca i32, align 4
771 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
772 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
773 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
774 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
775 // CHECK9-NEXT: [[I13:%.*]] = alloca i32, align 4
776 // CHECK9-NEXT: [[J14:%.*]] = alloca i32, align 4
777 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
778 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
779 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
780 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
781 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
782 // CHECK9-NEXT: store i64 [[VLA1]], i64* [[VLA_ADDR2]], align 8
783 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
784 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
785 // CHECK9-NEXT: [[CONV3:%.*]] = bitcast i64* [[M_ADDR]] to i32*
786 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
787 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR2]], align 8
788 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
789 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 4
790 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
791 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[CONV3]], align 4
792 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_5]], align 4
793 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
794 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
795 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
796 // CHECK9-NEXT: [[CONV7:%.*]] = sext i32 [[DIV]] to i64
797 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
798 // CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP6]], 0
799 // CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
800 // CHECK9-NEXT: [[CONV10:%.*]] = sext i32 [[DIV9]] to i64
801 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV7]], [[CONV10]]
802 // CHECK9-NEXT: [[SUB11:%.*]] = sub nsw i64 [[MUL]], 1
803 // CHECK9-NEXT: store i64 [[SUB11]], i64* [[DOTCAPTURE_EXPR_6]], align 8
804 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
805 // CHECK9-NEXT: store i32 0, i32* [[J]], align 4
806 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
807 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
808 // CHECK9-NEXT: br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
809 // CHECK9: land.lhs.true:
810 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
811 // CHECK9-NEXT: [[CMP12:%.*]] = icmp slt i32 0, [[TMP8]]
812 // CHECK9-NEXT: br i1 [[CMP12]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
813 // CHECK9: omp.precond.then:
814 // CHECK9-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
815 // CHECK9-NEXT: [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
816 // CHECK9-NEXT: store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
817 // CHECK9-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
818 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
819 // CHECK9-NEXT: [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
820 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
821 // CHECK9-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
822 // CHECK9-NEXT: [[TMP12:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
823 // CHECK9-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
824 // CHECK9-NEXT: [[CMP15:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
825 // CHECK9-NEXT: br i1 [[CMP15]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
826 // CHECK9: cond.true:
827 // CHECK9-NEXT: [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_6]], align 8
828 // CHECK9-NEXT: br label [[COND_END:%.*]]
829 // CHECK9: cond.false:
830 // CHECK9-NEXT: [[TMP15:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
831 // CHECK9-NEXT: br label [[COND_END]]
832 // CHECK9: cond.end:
833 // CHECK9-NEXT: [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
834 // CHECK9-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
835 // CHECK9-NEXT: [[TMP16:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
836 // CHECK9-NEXT: store i64 [[TMP16]], i64* [[DOTOMP_IV]], align 8
837 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
838 // CHECK9: omp.inner.for.cond:
839 // CHECK9-NEXT: [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
840 // CHECK9-NEXT: [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
841 // CHECK9-NEXT: [[CMP16:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
842 // CHECK9-NEXT: br i1 [[CMP16]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
843 // CHECK9: omp.inner.for.body:
844 // CHECK9-NEXT: [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
845 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4, !llvm.access.group !5
846 // CHECK9-NEXT: [[SUB17:%.*]] = sub nsw i32 [[TMP20]], 0
847 // CHECK9-NEXT: [[DIV18:%.*]] = sdiv i32 [[SUB17]], 1
848 // CHECK9-NEXT: [[MUL19:%.*]] = mul nsw i32 1, [[DIV18]]
849 // CHECK9-NEXT: [[CONV20:%.*]] = sext i32 [[MUL19]] to i64
850 // CHECK9-NEXT: [[DIV21:%.*]] = sdiv i64 [[TMP19]], [[CONV20]]
851 // CHECK9-NEXT: [[MUL22:%.*]] = mul nsw i64 [[DIV21]], 1
852 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL22]]
853 // CHECK9-NEXT: [[CONV23:%.*]] = trunc i64 [[ADD]] to i32
854 // CHECK9-NEXT: store i32 [[CONV23]], i32* [[I13]], align 4, !llvm.access.group !5
855 // CHECK9-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
856 // CHECK9-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
857 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4, !llvm.access.group !5
858 // CHECK9-NEXT: [[SUB24:%.*]] = sub nsw i32 [[TMP23]], 0
859 // CHECK9-NEXT: [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
860 // CHECK9-NEXT: [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
861 // CHECK9-NEXT: [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
862 // CHECK9-NEXT: [[DIV28:%.*]] = sdiv i64 [[TMP22]], [[CONV27]]
863 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4, !llvm.access.group !5
864 // CHECK9-NEXT: [[SUB29:%.*]] = sub nsw i32 [[TMP24]], 0
865 // CHECK9-NEXT: [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
866 // CHECK9-NEXT: [[MUL31:%.*]] = mul nsw i32 1, [[DIV30]]
867 // CHECK9-NEXT: [[CONV32:%.*]] = sext i32 [[MUL31]] to i64
868 // CHECK9-NEXT: [[MUL33:%.*]] = mul nsw i64 [[DIV28]], [[CONV32]]
869 // CHECK9-NEXT: [[SUB34:%.*]] = sub nsw i64 [[TMP21]], [[MUL33]]
870 // CHECK9-NEXT: [[MUL35:%.*]] = mul nsw i64 [[SUB34]], 1
871 // CHECK9-NEXT: [[ADD36:%.*]] = add nsw i64 0, [[MUL35]]
872 // CHECK9-NEXT: [[CONV37:%.*]] = trunc i64 [[ADD36]] to i32
873 // CHECK9-NEXT: store i32 [[CONV37]], i32* [[J14]], align 4, !llvm.access.group !5
874 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[I13]], align 4, !llvm.access.group !5
875 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP25]] to i64
876 // CHECK9-NEXT: [[TMP26:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP1]]
877 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[TMP26]]
878 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[J14]], align 4, !llvm.access.group !5
879 // CHECK9-NEXT: [[IDXPROM38:%.*]] = sext i32 [[TMP27]] to i64
880 // CHECK9-NEXT: [[ARRAYIDX39:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 [[IDXPROM38]]
881 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX39]], align 4, !llvm.access.group !5
882 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
883 // CHECK9: omp.body.continue:
884 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
885 // CHECK9: omp.inner.for.inc:
886 // CHECK9-NEXT: [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
887 // CHECK9-NEXT: [[ADD40:%.*]] = add nsw i64 [[TMP28]], 1
888 // CHECK9-NEXT: store i64 [[ADD40]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
889 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
890 // CHECK9: omp.inner.for.end:
891 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
892 // CHECK9: omp.loop.exit:
893 // CHECK9-NEXT: [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
894 // CHECK9-NEXT: [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
895 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
896 // CHECK9-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
897 // CHECK9-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
898 // CHECK9-NEXT: br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
899 // CHECK9: .omp.final.then:
900 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
901 // CHECK9-NEXT: [[SUB41:%.*]] = sub nsw i32 [[TMP33]], 0
902 // CHECK9-NEXT: [[DIV42:%.*]] = sdiv i32 [[SUB41]], 1
903 // CHECK9-NEXT: [[MUL43:%.*]] = mul nsw i32 [[DIV42]], 1
904 // CHECK9-NEXT: [[ADD44:%.*]] = add nsw i32 0, [[MUL43]]
905 // CHECK9-NEXT: store i32 [[ADD44]], i32* [[I13]], align 4
906 // CHECK9-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_5]], align 4
907 // CHECK9-NEXT: [[SUB45:%.*]] = sub nsw i32 [[TMP34]], 0
908 // CHECK9-NEXT: [[DIV46:%.*]] = sdiv i32 [[SUB45]], 1
909 // CHECK9-NEXT: [[MUL47:%.*]] = mul nsw i32 [[DIV46]], 1
910 // CHECK9-NEXT: [[ADD48:%.*]] = add nsw i32 0, [[MUL47]]
911 // CHECK9-NEXT: store i32 [[ADD48]], i32* [[J14]], align 4
912 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
913 // CHECK9: .omp.final.done:
914 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
915 // CHECK9: omp.precond.end:
916 // CHECK9-NEXT: ret void
917 //
918 //
919 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
920 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
921 // CHECK9-NEXT: entry:
922 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
923 // CHECK9-NEXT: [[A:%.*]] = alloca [10 x [2 x i32]], align 4
924 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
925 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
926 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
927 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
928 // CHECK9-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
929 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
930 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
931 // CHECK9-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x [2 x i32]]**
932 // CHECK9-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP1]], align 8
933 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
934 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x [2 x i32]]**
935 // CHECK9-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP3]], align 8
936 // CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
937 // CHECK9-NEXT: store i8* null, i8** [[TMP4]], align 8
938 // CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
939 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
940 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
941 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
942 // CHECK9-NEXT: store i32 1, i32* [[TMP7]], align 4
943 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
944 // CHECK9-NEXT: store i32 1, i32* [[TMP8]], align 4
945 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
946 // CHECK9-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
947 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
948 // CHECK9-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
949 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
950 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP11]], align 8
951 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
952 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP12]], align 8
953 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
954 // CHECK9-NEXT: store i8** null, i8*** [[TMP13]], align 8
955 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
956 // CHECK9-NEXT: store i8** null, i8*** [[TMP14]], align 8
957 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
958 // CHECK9-NEXT: store i64 20, i64* [[TMP15]], align 8
959 // CHECK9-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
960 // CHECK9-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
961 // CHECK9-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
962 // CHECK9: omp_offload.failed:
963 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67([10 x [2 x i32]]* [[A]]) #[[ATTR4]]
964 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
965 // CHECK9: omp_offload.cont:
966 // CHECK9-NEXT: ret i32 0
967 //
968 //
969 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67
970 // CHECK9-SAME: ([10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
971 // CHECK9-NEXT: entry:
972 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 8
973 // CHECK9-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 8
974 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 8
975 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x [2 x i32]]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x [2 x i32]]* [[TMP0]])
976 // CHECK9-NEXT: ret void
977 //
978 //
979 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
980 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR3]] {
981 // CHECK9-NEXT: entry:
982 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
983 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
984 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 8
985 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
986 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
987 // CHECK9-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
988 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
989 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
990 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
991 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
992 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
993 // CHECK9-NEXT: [[J:%.*]] = alloca i32, align 4
994 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
995 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
996 // CHECK9-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 8
997 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 8
998 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
999 // CHECK9-NEXT: store i32 19, i32* [[DOTOMP_UB]], align 4
1000 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1001 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1002 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1003 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1004 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1005 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1006 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1007 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1008 // CHECK9: cond.true:
1009 // CHECK9-NEXT: br label [[COND_END:%.*]]
1010 // CHECK9: cond.false:
1011 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1012 // CHECK9-NEXT: br label [[COND_END]]
1013 // CHECK9: cond.end:
1014 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1015 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1016 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1017 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1018 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1019 // CHECK9: omp.inner.for.cond:
1020 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1021 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
1022 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1023 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1024 // CHECK9: omp.inner.for.body:
1025 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1026 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
1027 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1028 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1029 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
1030 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1031 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1032 // CHECK9-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
1033 // CHECK9-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
1034 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
1035 // CHECK9-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
1036 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
1037 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !11
1038 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
1039 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1040 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[TMP0]], i64 0, i64 [[IDXPROM]]
1041 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !11
1042 // CHECK9-NEXT: [[IDXPROM7:%.*]] = sext i32 [[TMP12]] to i64
1043 // CHECK9-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM7]]
1044 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX8]], align 4, !llvm.access.group !11
1045 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1046 // CHECK9: omp.body.continue:
1047 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1048 // CHECK9: omp.inner.for.inc:
1049 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1050 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP13]], 1
1051 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1052 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1053 // CHECK9: omp.inner.for.end:
1054 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1055 // CHECK9: omp.loop.exit:
1056 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1057 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1058 // CHECK9-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1059 // CHECK9-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1060 // CHECK9: .omp.final.then:
1061 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
1062 // CHECK9-NEXT: store i32 2, i32* [[J]], align 4
1063 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
1064 // CHECK9: .omp.final.done:
1065 // CHECK9-NEXT: ret void
1066 //
1067 //
1068 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1069 // CHECK9-SAME: () #[[ATTR7:[0-9]+]] {
1070 // CHECK9-NEXT: entry:
1071 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
1072 // CHECK9-NEXT: ret void
1073 //
1074 //
1075 // CHECK11-LABEL: define {{[^@]+}}@main
1076 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1077 // CHECK11-NEXT: entry:
1078 // CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1079 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1080 // CHECK11-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
1081 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
1082 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
1083 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1084 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1085 // CHECK11-NEXT: [[__VLA_EXPR1:%.*]] = alloca i32, align 4
1086 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1087 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
1088 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x i8*], align 4
1089 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x i8*], align 4
1090 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x i8*], align 4
1091 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [5 x i64], align 4
1092 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1093 // CHECK11-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1094 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1095 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1096 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
1097 // CHECK11-NEXT: store i32 0, i32* [[RETVAL]], align 4
1098 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1099 // CHECK11-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
1100 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
1101 // CHECK11-NEXT: store i32 2, i32* [[M]], align 4
1102 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1103 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[M]], align 4
1104 // CHECK11-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1105 // CHECK11-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
1106 // CHECK11-NEXT: [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1107 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
1108 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1109 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[__VLA_EXPR1]], align 4
1110 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[N]], align 4
1111 // CHECK11-NEXT: store i32 [[TMP4]], i32* [[N_CASTED]], align 4
1112 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_CASTED]], align 4
1113 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[M]], align 4
1114 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[M_CASTED]], align 4
1115 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[M_CASTED]], align 4
1116 // CHECK11-NEXT: [[TMP8:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1117 // CHECK11-NEXT: [[TMP9:%.*]] = mul nuw i32 [[TMP8]], 4
1118 // CHECK11-NEXT: [[TMP10:%.*]] = sext i32 [[TMP9]] to i64
1119 // CHECK11-NEXT: [[TMP11:%.*]] = bitcast [5 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1120 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP11]], i8* align 4 bitcast ([5 x i64]* @.offload_sizes to i8*), i32 40, i1 false)
1121 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1122 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
1123 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[TMP13]], align 4
1124 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1125 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1126 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[TMP15]], align 4
1127 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1128 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
1129 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1130 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
1131 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[TMP18]], align 4
1132 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1133 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
1134 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[TMP20]], align 4
1135 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1136 // CHECK11-NEXT: store i8* null, i8** [[TMP21]], align 4
1137 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1138 // CHECK11-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
1139 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP23]], align 4
1140 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1141 // CHECK11-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to i32*
1142 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP25]], align 4
1143 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1144 // CHECK11-NEXT: store i8* null, i8** [[TMP26]], align 4
1145 // CHECK11-NEXT: [[TMP27:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1146 // CHECK11-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to i32*
1147 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[TMP28]], align 4
1148 // CHECK11-NEXT: [[TMP29:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1149 // CHECK11-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to i32*
1150 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[TMP30]], align 4
1151 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
1152 // CHECK11-NEXT: store i8* null, i8** [[TMP31]], align 4
1153 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1154 // CHECK11-NEXT: [[TMP33:%.*]] = bitcast i8** [[TMP32]] to i32**
1155 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP33]], align 4
1156 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1157 // CHECK11-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to i32**
1158 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP35]], align 4
1159 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 4
1160 // CHECK11-NEXT: store i64 [[TMP10]], i64* [[TMP36]], align 4
1161 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
1162 // CHECK11-NEXT: store i8* null, i8** [[TMP37]], align 4
1163 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1164 // CHECK11-NEXT: [[TMP39:%.*]] = getelementptr inbounds [5 x i8*], [5 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1165 // CHECK11-NEXT: [[TMP40:%.*]] = getelementptr inbounds [5 x i64], [5 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1166 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
1167 // CHECK11-NEXT: store i32 [[TMP41]], i32* [[DOTCAPTURE_EXPR_]], align 4
1168 // CHECK11-NEXT: [[TMP42:%.*]] = load i32, i32* [[M]], align 4
1169 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[DOTCAPTURE_EXPR_2]], align 4
1170 // CHECK11-NEXT: [[TMP43:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1171 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP43]], 0
1172 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1173 // CHECK11-NEXT: [[CONV:%.*]] = sext i32 [[DIV]] to i64
1174 // CHECK11-NEXT: [[TMP44:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1175 // CHECK11-NEXT: [[SUB4:%.*]] = sub nsw i32 [[TMP44]], 0
1176 // CHECK11-NEXT: [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
1177 // CHECK11-NEXT: [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
1178 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
1179 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
1180 // CHECK11-NEXT: store i64 [[SUB7]], i64* [[DOTCAPTURE_EXPR_3]], align 8
1181 // CHECK11-NEXT: [[TMP45:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
1182 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i64 [[TMP45]], 1
1183 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1184 // CHECK11-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1185 // CHECK11-NEXT: store i32 1, i32* [[TMP46]], align 4
1186 // CHECK11-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1187 // CHECK11-NEXT: store i32 5, i32* [[TMP47]], align 4
1188 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1189 // CHECK11-NEXT: store i8** [[TMP38]], i8*** [[TMP48]], align 4
1190 // CHECK11-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1191 // CHECK11-NEXT: store i8** [[TMP39]], i8*** [[TMP49]], align 4
1192 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1193 // CHECK11-NEXT: store i64* [[TMP40]], i64** [[TMP50]], align 4
1194 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1195 // CHECK11-NEXT: store i64* getelementptr inbounds ([5 x i64], [5 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP51]], align 4
1196 // CHECK11-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1197 // CHECK11-NEXT: store i8** null, i8*** [[TMP52]], align 4
1198 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1199 // CHECK11-NEXT: store i8** null, i8*** [[TMP53]], align 4
1200 // CHECK11-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1201 // CHECK11-NEXT: store i64 [[ADD]], i64* [[TMP54]], align 8
1202 // CHECK11-NEXT: [[TMP55:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1203 // CHECK11-NEXT: [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
1204 // CHECK11-NEXT: br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1205 // CHECK11: omp_offload.failed:
1206 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80(i32 [[TMP5]], i32 [[TMP7]], i32 [[TMP0]], i32 [[TMP1]], i32* [[VLA]]) #[[ATTR4:[0-9]+]]
1207 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
1208 // CHECK11: omp_offload.cont:
1209 // CHECK11-NEXT: [[TMP57:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1210 // CHECK11-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP57]])
1211 // CHECK11-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1212 // CHECK11-NEXT: [[TMP58:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
1213 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP58]])
1214 // CHECK11-NEXT: [[TMP59:%.*]] = load i32, i32* [[RETVAL]], align 4
1215 // CHECK11-NEXT: ret i32 [[TMP59]]
1216 //
1217 //
1218 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l80
1219 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1220 // CHECK11-NEXT: entry:
1221 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1222 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
1223 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1224 // CHECK11-NEXT: [[VLA_ADDR2:%.*]] = alloca i32, align 4
1225 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1226 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1227 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
1228 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1229 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
1230 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1231 // CHECK11-NEXT: store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
1232 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1233 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1234 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
1235 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1236 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
1237 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[N_CASTED]], align 4
1238 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[N_CASTED]], align 4
1239 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[M_ADDR]], align 4
1240 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[M_CASTED]], align 4
1241 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[M_CASTED]], align 4
1242 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32 [[TMP4]], i32 [[TMP6]], i32 [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
1243 // CHECK11-NEXT: ret void
1244 //
1245 //
1246 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1247 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[N:%.*]], i32 noundef [[M:%.*]], i32 noundef [[VLA:%.*]], i32 noundef [[VLA1:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR3:[0-9]+]] {
1248 // CHECK11-NEXT: entry:
1249 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1250 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1251 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1252 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
1253 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1254 // CHECK11-NEXT: [[VLA_ADDR2:%.*]] = alloca i32, align 4
1255 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1256 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
1257 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1258 // CHECK11-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
1259 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1260 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_4:%.*]] = alloca i32, align 4
1261 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_5:%.*]] = alloca i64, align 8
1262 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
1263 // CHECK11-NEXT: [[J:%.*]] = alloca i32, align 4
1264 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
1265 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
1266 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1267 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1268 // CHECK11-NEXT: [[I11:%.*]] = alloca i32, align 4
1269 // CHECK11-NEXT: [[J12:%.*]] = alloca i32, align 4
1270 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1271 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1272 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1273 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
1274 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1275 // CHECK11-NEXT: store i32 [[VLA1]], i32* [[VLA_ADDR2]], align 4
1276 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1277 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1278 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR2]], align 4
1279 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1280 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_ADDR]], align 4
1281 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1282 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[M_ADDR]], align 4
1283 // CHECK11-NEXT: store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_4]], align 4
1284 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1285 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
1286 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1287 // CHECK11-NEXT: [[CONV:%.*]] = sext i32 [[DIV]] to i64
1288 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1289 // CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP6]], 0
1290 // CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
1291 // CHECK11-NEXT: [[CONV8:%.*]] = sext i32 [[DIV7]] to i64
1292 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV8]]
1293 // CHECK11-NEXT: [[SUB9:%.*]] = sub nsw i64 [[MUL]], 1
1294 // CHECK11-NEXT: store i64 [[SUB9]], i64* [[DOTCAPTURE_EXPR_5]], align 8
1295 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
1296 // CHECK11-NEXT: store i32 0, i32* [[J]], align 4
1297 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1298 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP7]]
1299 // CHECK11-NEXT: br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
1300 // CHECK11: land.lhs.true:
1301 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1302 // CHECK11-NEXT: [[CMP10:%.*]] = icmp slt i32 0, [[TMP8]]
1303 // CHECK11-NEXT: br i1 [[CMP10]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
1304 // CHECK11: omp.precond.then:
1305 // CHECK11-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
1306 // CHECK11-NEXT: [[TMP9:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
1307 // CHECK11-NEXT: store i64 [[TMP9]], i64* [[DOTOMP_UB]], align 8
1308 // CHECK11-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1309 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1310 // CHECK11-NEXT: [[TMP10:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1311 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[TMP10]], align 4
1312 // CHECK11-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
1313 // CHECK11-NEXT: [[TMP12:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1314 // CHECK11-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
1315 // CHECK11-NEXT: [[CMP13:%.*]] = icmp sgt i64 [[TMP12]], [[TMP13]]
1316 // CHECK11-NEXT: br i1 [[CMP13]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1317 // CHECK11: cond.true:
1318 // CHECK11-NEXT: [[TMP14:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_5]], align 8
1319 // CHECK11-NEXT: br label [[COND_END:%.*]]
1320 // CHECK11: cond.false:
1321 // CHECK11-NEXT: [[TMP15:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1322 // CHECK11-NEXT: br label [[COND_END]]
1323 // CHECK11: cond.end:
1324 // CHECK11-NEXT: [[COND:%.*]] = phi i64 [ [[TMP14]], [[COND_TRUE]] ], [ [[TMP15]], [[COND_FALSE]] ]
1325 // CHECK11-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
1326 // CHECK11-NEXT: [[TMP16:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1327 // CHECK11-NEXT: store i64 [[TMP16]], i64* [[DOTOMP_IV]], align 8
1328 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1329 // CHECK11: omp.inner.for.cond:
1330 // CHECK11-NEXT: [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1331 // CHECK11-NEXT: [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !6
1332 // CHECK11-NEXT: [[CMP14:%.*]] = icmp sle i64 [[TMP17]], [[TMP18]]
1333 // CHECK11-NEXT: br i1 [[CMP14]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1334 // CHECK11: omp.inner.for.body:
1335 // CHECK11-NEXT: [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1336 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4, !llvm.access.group !6
1337 // CHECK11-NEXT: [[SUB15:%.*]] = sub nsw i32 [[TMP20]], 0
1338 // CHECK11-NEXT: [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
1339 // CHECK11-NEXT: [[MUL17:%.*]] = mul nsw i32 1, [[DIV16]]
1340 // CHECK11-NEXT: [[CONV18:%.*]] = sext i32 [[MUL17]] to i64
1341 // CHECK11-NEXT: [[DIV19:%.*]] = sdiv i64 [[TMP19]], [[CONV18]]
1342 // CHECK11-NEXT: [[MUL20:%.*]] = mul nsw i64 [[DIV19]], 1
1343 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL20]]
1344 // CHECK11-NEXT: [[CONV21:%.*]] = trunc i64 [[ADD]] to i32
1345 // CHECK11-NEXT: store i32 [[CONV21]], i32* [[I11]], align 4, !llvm.access.group !6
1346 // CHECK11-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1347 // CHECK11-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1348 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4, !llvm.access.group !6
1349 // CHECK11-NEXT: [[SUB22:%.*]] = sub nsw i32 [[TMP23]], 0
1350 // CHECK11-NEXT: [[DIV23:%.*]] = sdiv i32 [[SUB22]], 1
1351 // CHECK11-NEXT: [[MUL24:%.*]] = mul nsw i32 1, [[DIV23]]
1352 // CHECK11-NEXT: [[CONV25:%.*]] = sext i32 [[MUL24]] to i64
1353 // CHECK11-NEXT: [[DIV26:%.*]] = sdiv i64 [[TMP22]], [[CONV25]]
1354 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4, !llvm.access.group !6
1355 // CHECK11-NEXT: [[SUB27:%.*]] = sub nsw i32 [[TMP24]], 0
1356 // CHECK11-NEXT: [[DIV28:%.*]] = sdiv i32 [[SUB27]], 1
1357 // CHECK11-NEXT: [[MUL29:%.*]] = mul nsw i32 1, [[DIV28]]
1358 // CHECK11-NEXT: [[CONV30:%.*]] = sext i32 [[MUL29]] to i64
1359 // CHECK11-NEXT: [[MUL31:%.*]] = mul nsw i64 [[DIV26]], [[CONV30]]
1360 // CHECK11-NEXT: [[SUB32:%.*]] = sub nsw i64 [[TMP21]], [[MUL31]]
1361 // CHECK11-NEXT: [[MUL33:%.*]] = mul nsw i64 [[SUB32]], 1
1362 // CHECK11-NEXT: [[ADD34:%.*]] = add nsw i64 0, [[MUL33]]
1363 // CHECK11-NEXT: [[CONV35:%.*]] = trunc i64 [[ADD34]] to i32
1364 // CHECK11-NEXT: store i32 [[CONV35]], i32* [[J12]], align 4, !llvm.access.group !6
1365 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[I11]], align 4, !llvm.access.group !6
1366 // CHECK11-NEXT: [[TMP26:%.*]] = mul nsw i32 [[TMP25]], [[TMP1]]
1367 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP26]]
1368 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[J12]], align 4, !llvm.access.group !6
1369 // CHECK11-NEXT: [[ARRAYIDX36:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i32 [[TMP27]]
1370 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX36]], align 4, !llvm.access.group !6
1371 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1372 // CHECK11: omp.body.continue:
1373 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1374 // CHECK11: omp.inner.for.inc:
1375 // CHECK11-NEXT: [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1376 // CHECK11-NEXT: [[ADD37:%.*]] = add nsw i64 [[TMP28]], 1
1377 // CHECK11-NEXT: store i64 [[ADD37]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !6
1378 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
1379 // CHECK11: omp.inner.for.end:
1380 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1381 // CHECK11: omp.loop.exit:
1382 // CHECK11-NEXT: [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1383 // CHECK11-NEXT: [[TMP30:%.*]] = load i32, i32* [[TMP29]], align 4
1384 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP30]])
1385 // CHECK11-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1386 // CHECK11-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
1387 // CHECK11-NEXT: br i1 [[TMP32]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1388 // CHECK11: .omp.final.then:
1389 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1390 // CHECK11-NEXT: [[SUB38:%.*]] = sub nsw i32 [[TMP33]], 0
1391 // CHECK11-NEXT: [[DIV39:%.*]] = sdiv i32 [[SUB38]], 1
1392 // CHECK11-NEXT: [[MUL40:%.*]] = mul nsw i32 [[DIV39]], 1
1393 // CHECK11-NEXT: [[ADD41:%.*]] = add nsw i32 0, [[MUL40]]
1394 // CHECK11-NEXT: store i32 [[ADD41]], i32* [[I11]], align 4
1395 // CHECK11-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_4]], align 4
1396 // CHECK11-NEXT: [[SUB42:%.*]] = sub nsw i32 [[TMP34]], 0
1397 // CHECK11-NEXT: [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
1398 // CHECK11-NEXT: [[MUL44:%.*]] = mul nsw i32 [[DIV43]], 1
1399 // CHECK11-NEXT: [[ADD45:%.*]] = add nsw i32 0, [[MUL44]]
1400 // CHECK11-NEXT: store i32 [[ADD45]], i32* [[J12]], align 4
1401 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
1402 // CHECK11: .omp.final.done:
1403 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
1404 // CHECK11: omp.precond.end:
1405 // CHECK11-NEXT: ret void
1406 //
1407 //
1408 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1409 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR6:[0-9]+]] comdat {
1410 // CHECK11-NEXT: entry:
1411 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1412 // CHECK11-NEXT: [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1413 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1414 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1415 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1416 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1417 // CHECK11-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1418 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1419 // CHECK11-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1420 // CHECK11-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x [2 x i32]]**
1421 // CHECK11-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP1]], align 4
1422 // CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1423 // CHECK11-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x [2 x i32]]**
1424 // CHECK11-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[TMP3]], align 4
1425 // CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1426 // CHECK11-NEXT: store i8* null, i8** [[TMP4]], align 4
1427 // CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1428 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1429 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1430 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1431 // CHECK11-NEXT: store i32 1, i32* [[TMP7]], align 4
1432 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1433 // CHECK11-NEXT: store i32 1, i32* [[TMP8]], align 4
1434 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1435 // CHECK11-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
1436 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1437 // CHECK11-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
1438 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1439 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP11]], align 4
1440 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1441 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP12]], align 4
1442 // CHECK11-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1443 // CHECK11-NEXT: store i8** null, i8*** [[TMP13]], align 4
1444 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1445 // CHECK11-NEXT: store i8** null, i8*** [[TMP14]], align 4
1446 // CHECK11-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1447 // CHECK11-NEXT: store i64 20, i64* [[TMP15]], align 8
1448 // CHECK11-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1449 // CHECK11-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1450 // CHECK11-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1451 // CHECK11: omp_offload.failed:
1452 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67([10 x [2 x i32]]* [[A]]) #[[ATTR4]]
1453 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
1454 // CHECK11: omp_offload.cont:
1455 // CHECK11-NEXT: ret i32 0
1456 //
1457 //
1458 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10ELi2EEiT__l67
1459 // CHECK11-SAME: ([10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR2]] {
1460 // CHECK11-NEXT: entry:
1461 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 4
1462 // CHECK11-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 4
1463 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 4
1464 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x [2 x i32]]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x [2 x i32]]* [[TMP0]])
1465 // CHECK11-NEXT: ret void
1466 //
1467 //
1468 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
1469 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x [2 x i32]]* noundef nonnull align 4 dereferenceable(80) [[A:%.*]]) #[[ATTR3]] {
1470 // CHECK11-NEXT: entry:
1471 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1472 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1473 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x [2 x i32]]*, align 4
1474 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1475 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1476 // CHECK11-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1477 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1478 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1479 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1480 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1481 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
1482 // CHECK11-NEXT: [[J:%.*]] = alloca i32, align 4
1483 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1484 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1485 // CHECK11-NEXT: store [10 x [2 x i32]]* [[A]], [10 x [2 x i32]]** [[A_ADDR]], align 4
1486 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x [2 x i32]]*, [10 x [2 x i32]]** [[A_ADDR]], align 4
1487 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1488 // CHECK11-NEXT: store i32 19, i32* [[DOTOMP_UB]], align 4
1489 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1490 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1491 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1492 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1493 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1494 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1495 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 19
1496 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1497 // CHECK11: cond.true:
1498 // CHECK11-NEXT: br label [[COND_END:%.*]]
1499 // CHECK11: cond.false:
1500 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1501 // CHECK11-NEXT: br label [[COND_END]]
1502 // CHECK11: cond.end:
1503 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 19, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1504 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1505 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1506 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1507 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1508 // CHECK11: omp.inner.for.cond:
1509 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1510 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
1511 // CHECK11-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1512 // CHECK11-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1513 // CHECK11: omp.inner.for.body:
1514 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1515 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP8]], 2
1516 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1517 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1518 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
1519 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1520 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1521 // CHECK11-NEXT: [[DIV3:%.*]] = sdiv i32 [[TMP10]], 2
1522 // CHECK11-NEXT: [[MUL4:%.*]] = mul nsw i32 [[DIV3]], 2
1523 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP9]], [[MUL4]]
1524 // CHECK11-NEXT: [[MUL5:%.*]] = mul nsw i32 [[SUB]], 1
1525 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 0, [[MUL5]]
1526 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[J]], align 4, !llvm.access.group !12
1527 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
1528 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[TMP0]], i32 0, i32 [[TMP11]]
1529 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !12
1530 // CHECK11-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP12]]
1531 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX7]], align 4, !llvm.access.group !12
1532 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1533 // CHECK11: omp.body.continue:
1534 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1535 // CHECK11: omp.inner.for.inc:
1536 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1537 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP13]], 1
1538 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
1539 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1540 // CHECK11: omp.inner.for.end:
1541 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1542 // CHECK11: omp.loop.exit:
1543 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1544 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1545 // CHECK11-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1546 // CHECK11-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1547 // CHECK11: .omp.final.then:
1548 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
1549 // CHECK11-NEXT: store i32 2, i32* [[J]], align 4
1550 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
1551 // CHECK11: .omp.final.done:
1552 // CHECK11-NEXT: ret void
1553 //
1554 //
1555 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1556 // CHECK11-SAME: () #[[ATTR7:[0-9]+]] {
1557 // CHECK11-NEXT: entry:
1558 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
1559 // CHECK11-NEXT: ret void
1560 //
1561 //
1562 // CHECK13-LABEL: define {{[^@]+}}@main
1563 // CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1564 // CHECK13-NEXT: entry:
1565 // CHECK13-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1566 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1567 // CHECK13-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1568 // CHECK13-NEXT: [[N:%.*]] = alloca i32, align 4
1569 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
1570 // CHECK13-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1571 // CHECK13-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1572 // CHECK13-NEXT: [[__VLA_EXPR1:%.*]] = alloca i64, align 8
1573 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
1574 // CHECK13-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1575 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1576 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1577 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
1578 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
1579 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
1580 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
1581 // CHECK13-NEXT: [[J:%.*]] = alloca i32, align 4
1582 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
1583 // CHECK13-NEXT: [[I9:%.*]] = alloca i32, align 4
1584 // CHECK13-NEXT: [[J10:%.*]] = alloca i32, align 4
1585 // CHECK13-NEXT: store i32 0, i32* [[RETVAL]], align 4
1586 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1587 // CHECK13-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1588 // CHECK13-NEXT: store i32 100, i32* [[N]], align 4
1589 // CHECK13-NEXT: store i32 2, i32* [[M]], align 4
1590 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1591 // CHECK13-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1592 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[M]], align 4
1593 // CHECK13-NEXT: [[TMP3:%.*]] = zext i32 [[TMP2]] to i64
1594 // CHECK13-NEXT: [[TMP4:%.*]] = call i8* @llvm.stacksave()
1595 // CHECK13-NEXT: store i8* [[TMP4]], i8** [[SAVED_STACK]], align 8
1596 // CHECK13-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], [[TMP3]]
1597 // CHECK13-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP5]], align 4
1598 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1599 // CHECK13-NEXT: store i64 [[TMP3]], i64* [[__VLA_EXPR1]], align 8
1600 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[N]], align 4
1601 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTCAPTURE_EXPR_]], align 4
1602 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[M]], align 4
1603 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTCAPTURE_EXPR_2]], align 4
1604 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1605 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP8]], 0
1606 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1607 // CHECK13-NEXT: [[CONV:%.*]] = sext i32 [[DIV]] to i64
1608 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1609 // CHECK13-NEXT: [[SUB4:%.*]] = sub nsw i32 [[TMP9]], 0
1610 // CHECK13-NEXT: [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
1611 // CHECK13-NEXT: [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
1612 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
1613 // CHECK13-NEXT: [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
1614 // CHECK13-NEXT: store i64 [[SUB7]], i64* [[DOTCAPTURE_EXPR_3]], align 8
1615 // CHECK13-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
1616 // CHECK13-NEXT: [[TMP10:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
1617 // CHECK13-NEXT: store i64 [[TMP10]], i64* [[DOTOMP_UB]], align 8
1618 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
1619 // CHECK13-NEXT: store i32 0, i32* [[J]], align 4
1620 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1621 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP11]]
1622 // CHECK13-NEXT: br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[SIMD_IF_END:%.*]]
1623 // CHECK13: land.lhs.true:
1624 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1625 // CHECK13-NEXT: [[CMP8:%.*]] = icmp slt i32 0, [[TMP12]]
1626 // CHECK13-NEXT: br i1 [[CMP8]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END]]
1627 // CHECK13: simd.if.then:
1628 // CHECK13-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1629 // CHECK13-NEXT: store i64 [[TMP13]], i64* [[DOTOMP_IV]], align 8
1630 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1631 // CHECK13: omp.inner.for.cond:
1632 // CHECK13-NEXT: [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1633 // CHECK13-NEXT: [[TMP15:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !2
1634 // CHECK13-NEXT: [[CMP11:%.*]] = icmp sle i64 [[TMP14]], [[TMP15]]
1635 // CHECK13-NEXT: br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1636 // CHECK13: omp.inner.for.body:
1637 // CHECK13-NEXT: [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1638 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !2
1639 // CHECK13-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP17]], 0
1640 // CHECK13-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
1641 // CHECK13-NEXT: [[MUL14:%.*]] = mul nsw i32 1, [[DIV13]]
1642 // CHECK13-NEXT: [[CONV15:%.*]] = sext i32 [[MUL14]] to i64
1643 // CHECK13-NEXT: [[DIV16:%.*]] = sdiv i64 [[TMP16]], [[CONV15]]
1644 // CHECK13-NEXT: [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 1
1645 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL17]]
1646 // CHECK13-NEXT: [[CONV18:%.*]] = trunc i64 [[ADD]] to i32
1647 // CHECK13-NEXT: store i32 [[CONV18]], i32* [[I9]], align 4, !llvm.access.group !2
1648 // CHECK13-NEXT: [[TMP18:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1649 // CHECK13-NEXT: [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1650 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !2
1651 // CHECK13-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP20]], 0
1652 // CHECK13-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1
1653 // CHECK13-NEXT: [[MUL21:%.*]] = mul nsw i32 1, [[DIV20]]
1654 // CHECK13-NEXT: [[CONV22:%.*]] = sext i32 [[MUL21]] to i64
1655 // CHECK13-NEXT: [[DIV23:%.*]] = sdiv i64 [[TMP19]], [[CONV22]]
1656 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !2
1657 // CHECK13-NEXT: [[SUB24:%.*]] = sub nsw i32 [[TMP21]], 0
1658 // CHECK13-NEXT: [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
1659 // CHECK13-NEXT: [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
1660 // CHECK13-NEXT: [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
1661 // CHECK13-NEXT: [[MUL28:%.*]] = mul nsw i64 [[DIV23]], [[CONV27]]
1662 // CHECK13-NEXT: [[SUB29:%.*]] = sub nsw i64 [[TMP18]], [[MUL28]]
1663 // CHECK13-NEXT: [[MUL30:%.*]] = mul nsw i64 [[SUB29]], 1
1664 // CHECK13-NEXT: [[ADD31:%.*]] = add nsw i64 0, [[MUL30]]
1665 // CHECK13-NEXT: [[CONV32:%.*]] = trunc i64 [[ADD31]] to i32
1666 // CHECK13-NEXT: store i32 [[CONV32]], i32* [[J10]], align 4, !llvm.access.group !2
1667 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[I9]], align 4, !llvm.access.group !2
1668 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP22]] to i64
1669 // CHECK13-NEXT: [[TMP23:%.*]] = mul nsw i64 [[IDXPROM]], [[TMP3]]
1670 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[TMP23]]
1671 // CHECK13-NEXT: [[TMP24:%.*]] = load i32, i32* [[J10]], align 4, !llvm.access.group !2
1672 // CHECK13-NEXT: [[IDXPROM33:%.*]] = sext i32 [[TMP24]] to i64
1673 // CHECK13-NEXT: [[ARRAYIDX34:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i64 [[IDXPROM33]]
1674 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX34]], align 4, !llvm.access.group !2
1675 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1676 // CHECK13: omp.body.continue:
1677 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1678 // CHECK13: omp.inner.for.inc:
1679 // CHECK13-NEXT: [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1680 // CHECK13-NEXT: [[ADD35:%.*]] = add nsw i64 [[TMP25]], 1
1681 // CHECK13-NEXT: store i64 [[ADD35]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !2
1682 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1683 // CHECK13: omp.inner.for.end:
1684 // CHECK13-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1685 // CHECK13-NEXT: [[SUB36:%.*]] = sub nsw i32 [[TMP26]], 0
1686 // CHECK13-NEXT: [[DIV37:%.*]] = sdiv i32 [[SUB36]], 1
1687 // CHECK13-NEXT: [[MUL38:%.*]] = mul nsw i32 [[DIV37]], 1
1688 // CHECK13-NEXT: [[ADD39:%.*]] = add nsw i32 0, [[MUL38]]
1689 // CHECK13-NEXT: store i32 [[ADD39]], i32* [[I9]], align 4
1690 // CHECK13-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1691 // CHECK13-NEXT: [[SUB40:%.*]] = sub nsw i32 [[TMP27]], 0
1692 // CHECK13-NEXT: [[DIV41:%.*]] = sdiv i32 [[SUB40]], 1
1693 // CHECK13-NEXT: [[MUL42:%.*]] = mul nsw i32 [[DIV41]], 1
1694 // CHECK13-NEXT: [[ADD43:%.*]] = add nsw i32 0, [[MUL42]]
1695 // CHECK13-NEXT: store i32 [[ADD43]], i32* [[J10]], align 4
1696 // CHECK13-NEXT: br label [[SIMD_IF_END]]
1697 // CHECK13: simd.if.end:
1698 // CHECK13-NEXT: [[TMP28:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1699 // CHECK13-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef signext [[TMP28]])
1700 // CHECK13-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1701 // CHECK13-NEXT: [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1702 // CHECK13-NEXT: call void @llvm.stackrestore(i8* [[TMP29]])
1703 // CHECK13-NEXT: [[TMP30:%.*]] = load i32, i32* [[RETVAL]], align 4
1704 // CHECK13-NEXT: ret i32 [[TMP30]]
1705 //
1706 //
1707 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1708 // CHECK13-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
1709 // CHECK13-NEXT: entry:
1710 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1711 // CHECK13-NEXT: [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1712 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
1713 // CHECK13-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1714 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1715 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1716 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1717 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
1718 // CHECK13-NEXT: [[J:%.*]] = alloca i32, align 4
1719 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1720 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1721 // CHECK13-NEXT: store i32 19, i32* [[DOTOMP_UB]], align 4
1722 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1723 // CHECK13-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1724 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1725 // CHECK13: omp.inner.for.cond:
1726 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1727 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
1728 // CHECK13-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1729 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1730 // CHECK13: omp.inner.for.body:
1731 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1732 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP3]], 2
1733 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1734 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1735 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
1736 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1737 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1738 // CHECK13-NEXT: [[DIV2:%.*]] = sdiv i32 [[TMP5]], 2
1739 // CHECK13-NEXT: [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 2
1740 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], [[MUL3]]
1741 // CHECK13-NEXT: [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
1742 // CHECK13-NEXT: [[ADD5:%.*]] = add nsw i32 0, [[MUL4]]
1743 // CHECK13-NEXT: store i32 [[ADD5]], i32* [[J]], align 4, !llvm.access.group !6
1744 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !6
1745 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP6]] to i64
1746 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[A]], i64 0, i64 [[IDXPROM]]
1747 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !6
1748 // CHECK13-NEXT: [[IDXPROM6:%.*]] = sext i32 [[TMP7]] to i64
1749 // CHECK13-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i64 0, i64 [[IDXPROM6]]
1750 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX7]], align 4, !llvm.access.group !6
1751 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1752 // CHECK13: omp.body.continue:
1753 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1754 // CHECK13: omp.inner.for.inc:
1755 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1756 // CHECK13-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP8]], 1
1757 // CHECK13-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
1758 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
1759 // CHECK13: omp.inner.for.end:
1760 // CHECK13-NEXT: store i32 10, i32* [[I]], align 4
1761 // CHECK13-NEXT: store i32 2, i32* [[J]], align 4
1762 // CHECK13-NEXT: ret i32 0
1763 //
1764 //
1765 // CHECK15-LABEL: define {{[^@]+}}@main
1766 // CHECK15-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1767 // CHECK15-NEXT: entry:
1768 // CHECK15-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1769 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1770 // CHECK15-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
1771 // CHECK15-NEXT: [[N:%.*]] = alloca i32, align 4
1772 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
1773 // CHECK15-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1774 // CHECK15-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1775 // CHECK15-NEXT: [[__VLA_EXPR1:%.*]] = alloca i32, align 4
1776 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
1777 // CHECK15-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1778 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1779 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1780 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
1781 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
1782 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
1783 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
1784 // CHECK15-NEXT: [[J:%.*]] = alloca i32, align 4
1785 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
1786 // CHECK15-NEXT: [[I9:%.*]] = alloca i32, align 4
1787 // CHECK15-NEXT: [[J10:%.*]] = alloca i32, align 4
1788 // CHECK15-NEXT: store i32 0, i32* [[RETVAL]], align 4
1789 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1790 // CHECK15-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
1791 // CHECK15-NEXT: store i32 100, i32* [[N]], align 4
1792 // CHECK15-NEXT: store i32 2, i32* [[M]], align 4
1793 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1794 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[M]], align 4
1795 // CHECK15-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1796 // CHECK15-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 4
1797 // CHECK15-NEXT: [[TMP3:%.*]] = mul nuw i32 [[TMP0]], [[TMP1]]
1798 // CHECK15-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP3]], align 4
1799 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1800 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[__VLA_EXPR1]], align 4
1801 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[N]], align 4
1802 // CHECK15-NEXT: store i32 [[TMP4]], i32* [[DOTCAPTURE_EXPR_]], align 4
1803 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[M]], align 4
1804 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTCAPTURE_EXPR_2]], align 4
1805 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1806 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP6]], 0
1807 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1808 // CHECK15-NEXT: [[CONV:%.*]] = sext i32 [[DIV]] to i64
1809 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1810 // CHECK15-NEXT: [[SUB4:%.*]] = sub nsw i32 [[TMP7]], 0
1811 // CHECK15-NEXT: [[DIV5:%.*]] = sdiv i32 [[SUB4]], 1
1812 // CHECK15-NEXT: [[CONV6:%.*]] = sext i32 [[DIV5]] to i64
1813 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i64 [[CONV]], [[CONV6]]
1814 // CHECK15-NEXT: [[SUB7:%.*]] = sub nsw i64 [[MUL]], 1
1815 // CHECK15-NEXT: store i64 [[SUB7]], i64* [[DOTCAPTURE_EXPR_3]], align 8
1816 // CHECK15-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
1817 // CHECK15-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
1818 // CHECK15-NEXT: store i64 [[TMP8]], i64* [[DOTOMP_UB]], align 8
1819 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
1820 // CHECK15-NEXT: store i32 0, i32* [[J]], align 4
1821 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1822 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP9]]
1823 // CHECK15-NEXT: br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[SIMD_IF_END:%.*]]
1824 // CHECK15: land.lhs.true:
1825 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1826 // CHECK15-NEXT: [[CMP8:%.*]] = icmp slt i32 0, [[TMP10]]
1827 // CHECK15-NEXT: br i1 [[CMP8]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END]]
1828 // CHECK15: simd.if.then:
1829 // CHECK15-NEXT: [[TMP11:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1830 // CHECK15-NEXT: store i64 [[TMP11]], i64* [[DOTOMP_IV]], align 8
1831 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1832 // CHECK15: omp.inner.for.cond:
1833 // CHECK15-NEXT: [[TMP12:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1834 // CHECK15-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !3
1835 // CHECK15-NEXT: [[CMP11:%.*]] = icmp sle i64 [[TMP12]], [[TMP13]]
1836 // CHECK15-NEXT: br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1837 // CHECK15: omp.inner.for.body:
1838 // CHECK15-NEXT: [[TMP14:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1839 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
1840 // CHECK15-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP15]], 0
1841 // CHECK15-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
1842 // CHECK15-NEXT: [[MUL14:%.*]] = mul nsw i32 1, [[DIV13]]
1843 // CHECK15-NEXT: [[CONV15:%.*]] = sext i32 [[MUL14]] to i64
1844 // CHECK15-NEXT: [[DIV16:%.*]] = sdiv i64 [[TMP14]], [[CONV15]]
1845 // CHECK15-NEXT: [[MUL17:%.*]] = mul nsw i64 [[DIV16]], 1
1846 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL17]]
1847 // CHECK15-NEXT: [[CONV18:%.*]] = trunc i64 [[ADD]] to i32
1848 // CHECK15-NEXT: store i32 [[CONV18]], i32* [[I9]], align 4, !llvm.access.group !3
1849 // CHECK15-NEXT: [[TMP16:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1850 // CHECK15-NEXT: [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1851 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
1852 // CHECK15-NEXT: [[SUB19:%.*]] = sub nsw i32 [[TMP18]], 0
1853 // CHECK15-NEXT: [[DIV20:%.*]] = sdiv i32 [[SUB19]], 1
1854 // CHECK15-NEXT: [[MUL21:%.*]] = mul nsw i32 1, [[DIV20]]
1855 // CHECK15-NEXT: [[CONV22:%.*]] = sext i32 [[MUL21]] to i64
1856 // CHECK15-NEXT: [[DIV23:%.*]] = sdiv i64 [[TMP17]], [[CONV22]]
1857 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4, !llvm.access.group !3
1858 // CHECK15-NEXT: [[SUB24:%.*]] = sub nsw i32 [[TMP19]], 0
1859 // CHECK15-NEXT: [[DIV25:%.*]] = sdiv i32 [[SUB24]], 1
1860 // CHECK15-NEXT: [[MUL26:%.*]] = mul nsw i32 1, [[DIV25]]
1861 // CHECK15-NEXT: [[CONV27:%.*]] = sext i32 [[MUL26]] to i64
1862 // CHECK15-NEXT: [[MUL28:%.*]] = mul nsw i64 [[DIV23]], [[CONV27]]
1863 // CHECK15-NEXT: [[SUB29:%.*]] = sub nsw i64 [[TMP16]], [[MUL28]]
1864 // CHECK15-NEXT: [[MUL30:%.*]] = mul nsw i64 [[SUB29]], 1
1865 // CHECK15-NEXT: [[ADD31:%.*]] = add nsw i64 0, [[MUL30]]
1866 // CHECK15-NEXT: [[CONV32:%.*]] = trunc i64 [[ADD31]] to i32
1867 // CHECK15-NEXT: store i32 [[CONV32]], i32* [[J10]], align 4, !llvm.access.group !3
1868 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[I9]], align 4, !llvm.access.group !3
1869 // CHECK15-NEXT: [[TMP21:%.*]] = mul nsw i32 [[TMP20]], [[TMP1]]
1870 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
1871 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[J10]], align 4, !llvm.access.group !3
1872 // CHECK15-NEXT: [[ARRAYIDX33:%.*]] = getelementptr inbounds i32, i32* [[ARRAYIDX]], i32 [[TMP22]]
1873 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX33]], align 4, !llvm.access.group !3
1874 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1875 // CHECK15: omp.body.continue:
1876 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1877 // CHECK15: omp.inner.for.inc:
1878 // CHECK15-NEXT: [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1879 // CHECK15-NEXT: [[ADD34:%.*]] = add nsw i64 [[TMP23]], 1
1880 // CHECK15-NEXT: store i64 [[ADD34]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !3
1881 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
1882 // CHECK15: omp.inner.for.end:
1883 // CHECK15-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1884 // CHECK15-NEXT: [[SUB35:%.*]] = sub nsw i32 [[TMP24]], 0
1885 // CHECK15-NEXT: [[DIV36:%.*]] = sdiv i32 [[SUB35]], 1
1886 // CHECK15-NEXT: [[MUL37:%.*]] = mul nsw i32 [[DIV36]], 1
1887 // CHECK15-NEXT: [[ADD38:%.*]] = add nsw i32 0, [[MUL37]]
1888 // CHECK15-NEXT: store i32 [[ADD38]], i32* [[I9]], align 4
1889 // CHECK15-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1890 // CHECK15-NEXT: [[SUB39:%.*]] = sub nsw i32 [[TMP25]], 0
1891 // CHECK15-NEXT: [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
1892 // CHECK15-NEXT: [[MUL41:%.*]] = mul nsw i32 [[DIV40]], 1
1893 // CHECK15-NEXT: [[ADD42:%.*]] = add nsw i32 0, [[MUL41]]
1894 // CHECK15-NEXT: store i32 [[ADD42]], i32* [[J10]], align 4
1895 // CHECK15-NEXT: br label [[SIMD_IF_END]]
1896 // CHECK15: simd.if.end:
1897 // CHECK15-NEXT: [[TMP26:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1898 // CHECK15-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10ELi2EEiT_(i32 noundef [[TMP26]])
1899 // CHECK15-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1900 // CHECK15-NEXT: [[TMP27:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
1901 // CHECK15-NEXT: call void @llvm.stackrestore(i8* [[TMP27]])
1902 // CHECK15-NEXT: [[TMP28:%.*]] = load i32, i32* [[RETVAL]], align 4
1903 // CHECK15-NEXT: ret i32 [[TMP28]]
1904 //
1905 //
1906 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10ELi2EEiT_
1907 // CHECK15-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
1908 // CHECK15-NEXT: entry:
1909 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1910 // CHECK15-NEXT: [[A:%.*]] = alloca [10 x [2 x i32]], align 4
1911 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
1912 // CHECK15-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1913 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1914 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1915 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1916 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
1917 // CHECK15-NEXT: [[J:%.*]] = alloca i32, align 4
1918 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1919 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1920 // CHECK15-NEXT: store i32 19, i32* [[DOTOMP_UB]], align 4
1921 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1922 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1923 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1924 // CHECK15: omp.inner.for.cond:
1925 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1926 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !7
1927 // CHECK15-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1928 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1929 // CHECK15: omp.inner.for.body:
1930 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1931 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP3]], 2
1932 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1933 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1934 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !7
1935 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1936 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1937 // CHECK15-NEXT: [[DIV2:%.*]] = sdiv i32 [[TMP5]], 2
1938 // CHECK15-NEXT: [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 2
1939 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], [[MUL3]]
1940 // CHECK15-NEXT: [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
1941 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 0, [[MUL4]]
1942 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[J]], align 4, !llvm.access.group !7
1943 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !7
1944 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x [2 x i32]], [10 x [2 x i32]]* [[A]], i32 0, i32 [[TMP6]]
1945 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[J]], align 4, !llvm.access.group !7
1946 // CHECK15-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[ARRAYIDX]], i32 0, i32 [[TMP7]]
1947 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX6]], align 4, !llvm.access.group !7
1948 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1949 // CHECK15: omp.body.continue:
1950 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1951 // CHECK15: omp.inner.for.inc:
1952 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1953 // CHECK15-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP8]], 1
1954 // CHECK15-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !7
1955 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
1956 // CHECK15: omp.inner.for.end:
1957 // CHECK15-NEXT: store i32 10, i32* [[I]], align 4
1958 // CHECK15-NEXT: store i32 2, i32* [[J]], align 4
1959 // CHECK15-NEXT: ret i32 0
1960 //
1961