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 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5
6 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9
10 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
11 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
12 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK6
13
14 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
15 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
16 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
17 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
18 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
19 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK10
20
21 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
22 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
23 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
24
25 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
26 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
27 // RUN: %clang_cc1 -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK14
28
29 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
30 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
31 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
32
33 // expected-no-diagnostics
34 #ifndef HEADER
35 #define HEADER
36
37 void fn1();
38 void fn2();
39 void fn3();
40 void fn4();
41 void fn5();
42 void fn6();
43
44 int Arg;
45
gtid_test()46 void gtid_test() {
47 #pragma omp target
48 #pragma omp teams
49 #pragma omp distribute parallel for
50 for(int i = 0 ; i < 100; i++) {}
51
52 #pragma omp target
53 #pragma omp teams
54 #pragma omp distribute parallel for if (parallel: false)
55 for(int i = 0 ; i < 100; i++) {
56 gtid_test();
57 }
58 }
59
60
61 template <typename T>
tmain(T Arg)62 int tmain(T Arg) {
63 #pragma omp target
64 #pragma omp teams
65 #pragma omp distribute parallel for if (true)
66 for(int i = 0 ; i < 100; i++) {
67 fn1();
68 }
69 #pragma omp target
70 #pragma omp teams
71 #pragma omp distribute parallel for if (false)
72 for(int i = 0 ; i < 100; i++) {
73 fn2();
74 }
75 #pragma omp target
76 #pragma omp teams
77 #pragma omp distribute parallel for if (parallel: Arg)
78 for(int i = 0 ; i < 100; i++) {
79 fn3();
80 }
81 return 0;
82 }
83
main()84 int main() {
85 #pragma omp target
86 #pragma omp teams
87 #pragma omp distribute parallel for if (true)
88 for(int i = 0 ; i < 100; i++) {
89
90
91 fn4();
92 }
93
94 #pragma omp target
95 #pragma omp teams
96 #pragma omp distribute parallel for if (false)
97 for(int i = 0 ; i < 100; i++) {
98
99
100 fn5();
101 }
102
103 #pragma omp target
104 #pragma omp teams
105 #pragma omp distribute parallel for if (Arg)
106 for(int i = 0 ; i < 100; i++) {
107
108
109 fn6();
110 }
111
112 return tmain(Arg);
113 }
114
115
116
117
118
119
120 // call void [[T_OUTLINE_FUN_3:@.+]](
121
122 #endif
123 // CHECK1-LABEL: define {{[^@]+}}@_Z9gtid_testv
124 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
125 // CHECK1-NEXT: entry:
126 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
127 // CHECK1-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
128 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
129 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
130 // CHECK1-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
131 // CHECK1-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
132 // CHECK1: omp_offload.failed:
133 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
134 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
135 // CHECK1: omp_offload.cont:
136 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
137 // CHECK1-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
138 // CHECK1-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
139 // CHECK1-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
140 // CHECK1: omp_offload.failed2:
141 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
142 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT3]]
143 // CHECK1: omp_offload.cont3:
144 // CHECK1-NEXT: ret void
145 //
146 //
147 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
148 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
149 // CHECK1-NEXT: entry:
150 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
151 // CHECK1-NEXT: ret void
152 //
153 //
154 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
155 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
156 // CHECK1-NEXT: entry:
157 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
158 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
159 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
160 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
163 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
164 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
167 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
168 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
169 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
170 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
171 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
172 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
173 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
174 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
175 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
176 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
177 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
178 // CHECK1: cond.true:
179 // CHECK1-NEXT: br label [[COND_END:%.*]]
180 // CHECK1: cond.false:
181 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
182 // CHECK1-NEXT: br label [[COND_END]]
183 // CHECK1: cond.end:
184 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
185 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
186 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
187 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
188 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
189 // CHECK1: omp.inner.for.cond:
190 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
191 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
192 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
193 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
194 // CHECK1: omp.inner.for.body:
195 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
196 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
197 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
198 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
199 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
200 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
201 // CHECK1: omp.inner.for.inc:
202 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
203 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
204 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
205 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
206 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
207 // CHECK1: omp.inner.for.end:
208 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
209 // CHECK1: omp.loop.exit:
210 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
211 // CHECK1-NEXT: ret void
212 //
213 //
214 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
215 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
216 // CHECK1-NEXT: entry:
217 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
218 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
219 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
220 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
221 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
222 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
223 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
224 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
225 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
226 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
227 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
228 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
229 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
230 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
231 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
232 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
233 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
234 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
235 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
236 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
237 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
238 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
239 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
240 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
241 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
242 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
243 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
244 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
245 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
246 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
247 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
248 // CHECK1: cond.true:
249 // CHECK1-NEXT: br label [[COND_END:%.*]]
250 // CHECK1: cond.false:
251 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
252 // CHECK1-NEXT: br label [[COND_END]]
253 // CHECK1: cond.end:
254 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
255 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
256 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
257 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
258 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
259 // CHECK1: omp.inner.for.cond:
260 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
261 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
262 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
263 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
264 // CHECK1: omp.inner.for.body:
265 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
266 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
267 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
268 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
269 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
270 // CHECK1: omp.body.continue:
271 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
272 // CHECK1: omp.inner.for.inc:
273 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
274 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
275 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
276 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
277 // CHECK1: omp.inner.for.end:
278 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
279 // CHECK1: omp.loop.exit:
280 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
281 // CHECK1-NEXT: ret void
282 //
283 //
284 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
285 // CHECK1-SAME: () #[[ATTR1]] {
286 // CHECK1-NEXT: entry:
287 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
288 // CHECK1-NEXT: ret void
289 //
290 //
291 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
292 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
293 // CHECK1-NEXT: entry:
294 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
295 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
296 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
297 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
299 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
300 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
301 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
302 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
303 // CHECK1-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
304 // CHECK1-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
305 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
306 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
307 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
308 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
309 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
310 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
311 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
312 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
313 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
314 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
315 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
316 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
317 // CHECK1: cond.true:
318 // CHECK1-NEXT: br label [[COND_END:%.*]]
319 // CHECK1: cond.false:
320 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
321 // CHECK1-NEXT: br label [[COND_END]]
322 // CHECK1: cond.end:
323 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
324 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
325 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
326 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
327 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
328 // CHECK1: omp.inner.for.cond:
329 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
330 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
331 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
332 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
333 // CHECK1: omp.inner.for.body:
334 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
335 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
336 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
337 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
338 // CHECK1-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
339 // CHECK1-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
340 // CHECK1-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
341 // CHECK1-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
342 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
343 // CHECK1: omp.inner.for.inc:
344 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
345 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
346 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
347 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
348 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
349 // CHECK1: omp.inner.for.end:
350 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
351 // CHECK1: omp.loop.exit:
352 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
353 // CHECK1-NEXT: ret void
354 //
355 //
356 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
357 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
358 // CHECK1-NEXT: entry:
359 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
360 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
361 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
362 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
363 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
364 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
365 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
366 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
367 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
368 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
369 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
370 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
371 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
372 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
373 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
374 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
375 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
376 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
377 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
378 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
379 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
380 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
381 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
382 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
383 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
384 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
385 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
386 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
387 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
388 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
389 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
390 // CHECK1: cond.true:
391 // CHECK1-NEXT: br label [[COND_END:%.*]]
392 // CHECK1: cond.false:
393 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
394 // CHECK1-NEXT: br label [[COND_END]]
395 // CHECK1: cond.end:
396 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
397 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
398 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
399 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
400 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
401 // CHECK1: omp.inner.for.cond:
402 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
403 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
404 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
405 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
406 // CHECK1: omp.inner.for.body:
407 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
408 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
409 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
410 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
411 // CHECK1-NEXT: call void @_Z9gtid_testv()
412 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
413 // CHECK1: omp.body.continue:
414 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
415 // CHECK1: omp.inner.for.inc:
416 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
417 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
418 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
419 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
420 // CHECK1: omp.inner.for.end:
421 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
422 // CHECK1: omp.loop.exit:
423 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
424 // CHECK1-NEXT: ret void
425 //
426 //
427 // CHECK1-LABEL: define {{[^@]+}}@main
428 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
429 // CHECK1-NEXT: entry:
430 // CHECK1-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
431 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
432 // CHECK1-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
433 // CHECK1-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
434 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
435 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
436 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
437 // CHECK1-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
438 // CHECK1-NEXT: store i32 0, i32* [[RETVAL]], align 4
439 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
440 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
441 // CHECK1-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
442 // CHECK1-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
443 // CHECK1: omp_offload.failed:
444 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
445 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
446 // CHECK1: omp_offload.cont:
447 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
448 // CHECK1-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
449 // CHECK1-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
450 // CHECK1-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
451 // CHECK1: omp_offload.failed2:
452 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
453 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT3]]
454 // CHECK1: omp_offload.cont3:
455 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
456 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
457 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
458 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
459 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
460 // CHECK1-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
461 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
462 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
463 // CHECK1-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
464 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
465 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
466 // CHECK1-NEXT: store i8* null, i8** [[TMP10]], align 8
467 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
468 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
469 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
470 // CHECK1-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
471 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
472 // CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
473 // CHECK1: omp_offload.failed5:
474 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
475 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT6]]
476 // CHECK1: omp_offload.cont6:
477 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
478 // CHECK1-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
479 // CHECK1-NEXT: ret i32 [[CALL]]
480 //
481 //
482 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
483 // CHECK1-SAME: () #[[ATTR1]] {
484 // CHECK1-NEXT: entry:
485 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
486 // CHECK1-NEXT: ret void
487 //
488 //
489 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
490 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
491 // CHECK1-NEXT: entry:
492 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
493 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
494 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
495 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
496 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
497 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
498 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
499 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
500 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
501 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
502 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
503 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
504 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
505 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
506 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
507 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
508 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
509 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
510 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
511 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
512 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
513 // CHECK1: cond.true:
514 // CHECK1-NEXT: br label [[COND_END:%.*]]
515 // CHECK1: cond.false:
516 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
517 // CHECK1-NEXT: br label [[COND_END]]
518 // CHECK1: cond.end:
519 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
520 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
521 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
522 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
523 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
524 // CHECK1: omp.inner.for.cond:
525 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
526 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
527 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
528 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
529 // CHECK1: omp.inner.for.body:
530 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
531 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
532 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
533 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
534 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
535 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
536 // CHECK1: omp.inner.for.inc:
537 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
538 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
539 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
540 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
541 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
542 // CHECK1: omp.inner.for.end:
543 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
544 // CHECK1: omp.loop.exit:
545 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
546 // CHECK1-NEXT: ret void
547 //
548 //
549 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
550 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
551 // CHECK1-NEXT: entry:
552 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
553 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
554 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
555 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
556 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
557 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
558 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
559 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
560 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
561 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
562 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
563 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
564 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
565 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
566 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
567 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
568 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
569 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
570 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
571 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
572 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
573 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
574 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
575 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
576 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
577 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
578 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
579 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
580 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
581 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
582 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
583 // CHECK1: cond.true:
584 // CHECK1-NEXT: br label [[COND_END:%.*]]
585 // CHECK1: cond.false:
586 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
587 // CHECK1-NEXT: br label [[COND_END]]
588 // CHECK1: cond.end:
589 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
590 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
591 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
592 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
593 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
594 // CHECK1: omp.inner.for.cond:
595 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
596 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
597 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
598 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
599 // CHECK1: omp.inner.for.body:
600 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
601 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
602 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
603 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
604 // CHECK1-NEXT: call void @_Z3fn4v()
605 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
606 // CHECK1: omp.body.continue:
607 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
608 // CHECK1: omp.inner.for.inc:
609 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
610 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
611 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
612 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
613 // CHECK1: omp.inner.for.end:
614 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
615 // CHECK1: omp.loop.exit:
616 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
617 // CHECK1-NEXT: ret void
618 //
619 //
620 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
621 // CHECK1-SAME: () #[[ATTR1]] {
622 // CHECK1-NEXT: entry:
623 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
624 // CHECK1-NEXT: ret void
625 //
626 //
627 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
628 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
629 // CHECK1-NEXT: entry:
630 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
631 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
632 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
633 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
634 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
635 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
636 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
637 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
638 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
639 // CHECK1-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
640 // CHECK1-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
641 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
642 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
643 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
644 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
645 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
646 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
647 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
648 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
649 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
650 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
651 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
652 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
653 // CHECK1: cond.true:
654 // CHECK1-NEXT: br label [[COND_END:%.*]]
655 // CHECK1: cond.false:
656 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
657 // CHECK1-NEXT: br label [[COND_END]]
658 // CHECK1: cond.end:
659 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
660 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
661 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
662 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
663 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
664 // CHECK1: omp.inner.for.cond:
665 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
666 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
667 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
668 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
669 // CHECK1: omp.inner.for.body:
670 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
671 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
672 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
673 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
674 // CHECK1-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
675 // CHECK1-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
676 // CHECK1-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
677 // CHECK1-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
678 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
679 // CHECK1: omp.inner.for.inc:
680 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
681 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
682 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
683 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
684 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
685 // CHECK1: omp.inner.for.end:
686 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
687 // CHECK1: omp.loop.exit:
688 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
689 // CHECK1-NEXT: ret void
690 //
691 //
692 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
693 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
694 // CHECK1-NEXT: entry:
695 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
696 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
697 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
698 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
699 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
700 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
701 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
702 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
703 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
704 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
705 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
706 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
707 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
708 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
709 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
710 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
711 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
712 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
713 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
714 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
715 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
716 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
717 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
718 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
719 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
720 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
721 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
722 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
723 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
724 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
725 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
726 // CHECK1: cond.true:
727 // CHECK1-NEXT: br label [[COND_END:%.*]]
728 // CHECK1: cond.false:
729 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
730 // CHECK1-NEXT: br label [[COND_END]]
731 // CHECK1: cond.end:
732 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
733 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
734 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
735 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
736 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
737 // CHECK1: omp.inner.for.cond:
738 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
739 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
740 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
741 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
742 // CHECK1: omp.inner.for.body:
743 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
744 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
745 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
746 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
747 // CHECK1-NEXT: call void @_Z3fn5v()
748 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
749 // CHECK1: omp.body.continue:
750 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
751 // CHECK1: omp.inner.for.inc:
752 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
753 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
754 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
755 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
756 // CHECK1: omp.inner.for.end:
757 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
758 // CHECK1: omp.loop.exit:
759 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
760 // CHECK1-NEXT: ret void
761 //
762 //
763 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
764 // CHECK1-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
765 // CHECK1-NEXT: entry:
766 // CHECK1-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
767 // CHECK1-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
768 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
769 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
770 // CHECK1-NEXT: ret void
771 //
772 //
773 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..8
774 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
775 // CHECK1-NEXT: entry:
776 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
777 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
778 // CHECK1-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
779 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
780 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
781 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
782 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
783 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
784 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
785 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
786 // CHECK1-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
787 // CHECK1-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
788 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
789 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
790 // CHECK1-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
791 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
792 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
793 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
794 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
795 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
796 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
797 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
798 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
799 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
800 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
801 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
802 // CHECK1: cond.true:
803 // CHECK1-NEXT: br label [[COND_END:%.*]]
804 // CHECK1: cond.false:
805 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
806 // CHECK1-NEXT: br label [[COND_END]]
807 // CHECK1: cond.end:
808 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
809 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
810 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
811 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
812 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
813 // CHECK1: omp.inner.for.cond:
814 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
815 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
816 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
817 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
818 // CHECK1: omp.inner.for.body:
819 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
820 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
821 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
822 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
823 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
824 // CHECK1-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
825 // CHECK1-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
826 // CHECK1: omp_if.then:
827 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
828 // CHECK1-NEXT: br label [[OMP_IF_END:%.*]]
829 // CHECK1: omp_if.else:
830 // CHECK1-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
831 // CHECK1-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
832 // CHECK1-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
833 // CHECK1-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
834 // CHECK1-NEXT: br label [[OMP_IF_END]]
835 // CHECK1: omp_if.end:
836 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
837 // CHECK1: omp.inner.for.inc:
838 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
839 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
840 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
841 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
842 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
843 // CHECK1: omp.inner.for.end:
844 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
845 // CHECK1: omp.loop.exit:
846 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
847 // CHECK1-NEXT: ret void
848 //
849 //
850 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
851 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
852 // CHECK1-NEXT: entry:
853 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
854 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
855 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
856 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
857 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
858 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
859 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
860 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
861 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
862 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
863 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
864 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
865 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
866 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
867 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
868 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
869 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
870 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
871 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
872 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
873 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
874 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
875 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
876 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
877 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
878 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
879 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
880 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
881 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
882 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
883 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
884 // CHECK1: cond.true:
885 // CHECK1-NEXT: br label [[COND_END:%.*]]
886 // CHECK1: cond.false:
887 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
888 // CHECK1-NEXT: br label [[COND_END]]
889 // CHECK1: cond.end:
890 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
891 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
892 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
893 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
894 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
895 // CHECK1: omp.inner.for.cond:
896 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
897 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
898 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
899 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
900 // CHECK1: omp.inner.for.body:
901 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
902 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
903 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
904 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
905 // CHECK1-NEXT: call void @_Z3fn6v()
906 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
907 // CHECK1: omp.body.continue:
908 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
909 // CHECK1: omp.inner.for.inc:
910 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
911 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
912 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
913 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
914 // CHECK1: omp.inner.for.end:
915 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
916 // CHECK1: omp.loop.exit:
917 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
918 // CHECK1-NEXT: ret void
919 //
920 //
921 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
922 // CHECK1-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
923 // CHECK1-NEXT: entry:
924 // CHECK1-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
925 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
926 // CHECK1-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
927 // CHECK1-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
928 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
929 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
930 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
931 // CHECK1-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
932 // CHECK1-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
933 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
934 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
935 // CHECK1-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
936 // CHECK1-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
937 // CHECK1: omp_offload.failed:
938 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
939 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
940 // CHECK1: omp_offload.cont:
941 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
942 // CHECK1-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
943 // CHECK1-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
944 // CHECK1-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
945 // CHECK1: omp_offload.failed2:
946 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
947 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT3]]
948 // CHECK1: omp_offload.cont3:
949 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
950 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
951 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
952 // CHECK1-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
953 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
954 // CHECK1-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
955 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
956 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
957 // CHECK1-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
958 // CHECK1-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
959 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
960 // CHECK1-NEXT: store i8* null, i8** [[TMP10]], align 8
961 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
962 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
963 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
964 // CHECK1-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
965 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
966 // CHECK1-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
967 // CHECK1: omp_offload.failed5:
968 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
969 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT6]]
970 // CHECK1: omp_offload.cont6:
971 // CHECK1-NEXT: ret i32 0
972 //
973 //
974 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
975 // CHECK1-SAME: () #[[ATTR1]] {
976 // CHECK1-NEXT: entry:
977 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
978 // CHECK1-NEXT: ret void
979 //
980 //
981 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
982 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
983 // CHECK1-NEXT: entry:
984 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
985 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
986 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
987 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
988 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
989 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
990 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
991 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
992 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
993 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
994 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
995 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
996 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
997 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
998 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
999 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1000 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1001 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1002 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1003 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1004 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1005 // CHECK1: cond.true:
1006 // CHECK1-NEXT: br label [[COND_END:%.*]]
1007 // CHECK1: cond.false:
1008 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1009 // CHECK1-NEXT: br label [[COND_END]]
1010 // CHECK1: cond.end:
1011 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1012 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1013 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1014 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1015 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1016 // CHECK1: omp.inner.for.cond:
1017 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1018 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1019 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1020 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1021 // CHECK1: omp.inner.for.body:
1022 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1023 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1024 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1025 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1026 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
1027 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1028 // CHECK1: omp.inner.for.inc:
1029 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1030 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1031 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1032 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1033 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1034 // CHECK1: omp.inner.for.end:
1035 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1036 // CHECK1: omp.loop.exit:
1037 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1038 // CHECK1-NEXT: ret void
1039 //
1040 //
1041 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
1042 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1043 // CHECK1-NEXT: entry:
1044 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1045 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1046 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1047 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1048 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1049 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1050 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1051 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1052 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1053 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1054 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1055 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1056 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1057 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1058 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1059 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1060 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1061 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1062 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1063 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1064 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1065 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1066 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1067 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1068 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1069 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1070 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1071 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1072 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1073 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1074 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1075 // CHECK1: cond.true:
1076 // CHECK1-NEXT: br label [[COND_END:%.*]]
1077 // CHECK1: cond.false:
1078 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1079 // CHECK1-NEXT: br label [[COND_END]]
1080 // CHECK1: cond.end:
1081 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1082 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1083 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1084 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1085 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1086 // CHECK1: omp.inner.for.cond:
1087 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1088 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1089 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1090 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1091 // CHECK1: omp.inner.for.body:
1092 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1093 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1094 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1095 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1096 // CHECK1-NEXT: call void @_Z3fn1v()
1097 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1098 // CHECK1: omp.body.continue:
1099 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1100 // CHECK1: omp.inner.for.inc:
1101 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1102 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1103 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1104 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1105 // CHECK1: omp.inner.for.end:
1106 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1107 // CHECK1: omp.loop.exit:
1108 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1109 // CHECK1-NEXT: ret void
1110 //
1111 //
1112 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
1113 // CHECK1-SAME: () #[[ATTR1]] {
1114 // CHECK1-NEXT: entry:
1115 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
1116 // CHECK1-NEXT: ret void
1117 //
1118 //
1119 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
1120 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1121 // CHECK1-NEXT: entry:
1122 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1123 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1124 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1125 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1126 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1127 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1128 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1129 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1130 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1131 // CHECK1-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1132 // CHECK1-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1133 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1134 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1135 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1136 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1137 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1138 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1139 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1140 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1141 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1142 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1143 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1144 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1145 // CHECK1: cond.true:
1146 // CHECK1-NEXT: br label [[COND_END:%.*]]
1147 // CHECK1: cond.false:
1148 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1149 // CHECK1-NEXT: br label [[COND_END]]
1150 // CHECK1: cond.end:
1151 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1152 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1153 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1154 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1155 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1156 // CHECK1: omp.inner.for.cond:
1157 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1158 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1159 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1160 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1161 // CHECK1: omp.inner.for.body:
1162 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1163 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1164 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1165 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1166 // CHECK1-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1167 // CHECK1-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1168 // CHECK1-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
1169 // CHECK1-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1170 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1171 // CHECK1: omp.inner.for.inc:
1172 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1173 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1174 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1175 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1176 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1177 // CHECK1: omp.inner.for.end:
1178 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1179 // CHECK1: omp.loop.exit:
1180 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1181 // CHECK1-NEXT: ret void
1182 //
1183 //
1184 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
1185 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1186 // CHECK1-NEXT: entry:
1187 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1188 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1189 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1190 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1191 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1192 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1193 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1194 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1195 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1196 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1197 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1198 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1199 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1200 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1201 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1202 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1203 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1204 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1205 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1206 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1207 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1208 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1209 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1210 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1211 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1212 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1213 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1214 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1215 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1216 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1217 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1218 // CHECK1: cond.true:
1219 // CHECK1-NEXT: br label [[COND_END:%.*]]
1220 // CHECK1: cond.false:
1221 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1222 // CHECK1-NEXT: br label [[COND_END]]
1223 // CHECK1: cond.end:
1224 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1225 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1226 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1227 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1228 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1229 // CHECK1: omp.inner.for.cond:
1230 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1231 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1232 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1233 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1234 // CHECK1: omp.inner.for.body:
1235 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1236 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1237 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1238 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1239 // CHECK1-NEXT: call void @_Z3fn2v()
1240 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1241 // CHECK1: omp.body.continue:
1242 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1243 // CHECK1: omp.inner.for.inc:
1244 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1245 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1246 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1247 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1248 // CHECK1: omp.inner.for.end:
1249 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1250 // CHECK1: omp.loop.exit:
1251 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1252 // CHECK1-NEXT: ret void
1253 //
1254 //
1255 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
1256 // CHECK1-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
1257 // CHECK1-NEXT: entry:
1258 // CHECK1-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
1259 // CHECK1-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
1260 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
1261 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
1262 // CHECK1-NEXT: ret void
1263 //
1264 //
1265 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
1266 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
1267 // CHECK1-NEXT: entry:
1268 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1269 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1270 // CHECK1-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
1271 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1272 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1273 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1274 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1275 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1276 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1277 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1278 // CHECK1-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1279 // CHECK1-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1280 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1281 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1282 // CHECK1-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
1283 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
1284 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1285 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1286 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1287 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1288 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1289 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1290 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1291 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1292 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
1293 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1294 // CHECK1: cond.true:
1295 // CHECK1-NEXT: br label [[COND_END:%.*]]
1296 // CHECK1: cond.false:
1297 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1298 // CHECK1-NEXT: br label [[COND_END]]
1299 // CHECK1: cond.end:
1300 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1301 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1302 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1303 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1304 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1305 // CHECK1: omp.inner.for.cond:
1306 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1307 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1308 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1309 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1310 // CHECK1: omp.inner.for.body:
1311 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1312 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1313 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1314 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1315 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
1316 // CHECK1-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
1317 // CHECK1-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1318 // CHECK1: omp_if.then:
1319 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
1320 // CHECK1-NEXT: br label [[OMP_IF_END:%.*]]
1321 // CHECK1: omp_if.else:
1322 // CHECK1-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
1323 // CHECK1-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1324 // CHECK1-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
1325 // CHECK1-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
1326 // CHECK1-NEXT: br label [[OMP_IF_END]]
1327 // CHECK1: omp_if.end:
1328 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1329 // CHECK1: omp.inner.for.inc:
1330 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1331 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1332 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
1333 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1334 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1335 // CHECK1: omp.inner.for.end:
1336 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1337 // CHECK1: omp.loop.exit:
1338 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1339 // CHECK1-NEXT: ret void
1340 //
1341 //
1342 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
1343 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1344 // CHECK1-NEXT: entry:
1345 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1346 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1347 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1348 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1349 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1350 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1351 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1352 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1353 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1354 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1355 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1356 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1357 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1358 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1359 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1360 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1361 // CHECK1-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1362 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1363 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1364 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1365 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1366 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1367 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1368 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1369 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1370 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1371 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1372 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1373 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1374 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1375 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1376 // CHECK1: cond.true:
1377 // CHECK1-NEXT: br label [[COND_END:%.*]]
1378 // CHECK1: cond.false:
1379 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1380 // CHECK1-NEXT: br label [[COND_END]]
1381 // CHECK1: cond.end:
1382 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1383 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1384 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1385 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1386 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1387 // CHECK1: omp.inner.for.cond:
1388 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1389 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1390 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1391 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1392 // CHECK1: omp.inner.for.body:
1393 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1394 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1395 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1396 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1397 // CHECK1-NEXT: call void @_Z3fn3v()
1398 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1399 // CHECK1: omp.body.continue:
1400 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1401 // CHECK1: omp.inner.for.inc:
1402 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1403 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1404 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1405 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1406 // CHECK1: omp.inner.for.end:
1407 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1408 // CHECK1: omp.loop.exit:
1409 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1410 // CHECK1-NEXT: ret void
1411 //
1412 //
1413 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1414 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
1415 // CHECK1-NEXT: entry:
1416 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
1417 // CHECK1-NEXT: ret void
1418 //
1419 //
1420 // CHECK2-LABEL: define {{[^@]+}}@_Z9gtid_testv
1421 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
1422 // CHECK2-NEXT: entry:
1423 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1424 // CHECK2-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1425 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
1426 // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
1427 // CHECK2-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1428 // CHECK2-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1429 // CHECK2: omp_offload.failed:
1430 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
1431 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
1432 // CHECK2: omp_offload.cont:
1433 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1434 // CHECK2-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
1435 // CHECK2-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1436 // CHECK2-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1437 // CHECK2: omp_offload.failed2:
1438 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
1439 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT3]]
1440 // CHECK2: omp_offload.cont3:
1441 // CHECK2-NEXT: ret void
1442 //
1443 //
1444 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
1445 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
1446 // CHECK2-NEXT: entry:
1447 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1448 // CHECK2-NEXT: ret void
1449 //
1450 //
1451 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1452 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1453 // CHECK2-NEXT: entry:
1454 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1455 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1456 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1457 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1458 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1459 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1460 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1461 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1462 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1463 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1464 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1465 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1466 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1467 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1468 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1469 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1470 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1471 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1472 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1473 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1474 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1475 // CHECK2: cond.true:
1476 // CHECK2-NEXT: br label [[COND_END:%.*]]
1477 // CHECK2: cond.false:
1478 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1479 // CHECK2-NEXT: br label [[COND_END]]
1480 // CHECK2: cond.end:
1481 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1482 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1483 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1484 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1485 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1486 // CHECK2: omp.inner.for.cond:
1487 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1488 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1489 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1490 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1491 // CHECK2: omp.inner.for.body:
1492 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1493 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1494 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1495 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1496 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
1497 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1498 // CHECK2: omp.inner.for.inc:
1499 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1500 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1501 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1502 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1503 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1504 // CHECK2: omp.inner.for.end:
1505 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1506 // CHECK2: omp.loop.exit:
1507 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1508 // CHECK2-NEXT: ret void
1509 //
1510 //
1511 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
1512 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1513 // CHECK2-NEXT: entry:
1514 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1515 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1516 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1517 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1518 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1519 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1520 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1521 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1522 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1523 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1524 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1525 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1526 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1527 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1528 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1529 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1530 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1531 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1532 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1533 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1534 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1535 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1536 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1537 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1538 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1539 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1540 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1541 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1542 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1543 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1544 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1545 // CHECK2: cond.true:
1546 // CHECK2-NEXT: br label [[COND_END:%.*]]
1547 // CHECK2: cond.false:
1548 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1549 // CHECK2-NEXT: br label [[COND_END]]
1550 // CHECK2: cond.end:
1551 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1552 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1553 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1554 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1555 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1556 // CHECK2: omp.inner.for.cond:
1557 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1558 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1559 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1560 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1561 // CHECK2: omp.inner.for.body:
1562 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1563 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1564 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1565 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1566 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1567 // CHECK2: omp.body.continue:
1568 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1569 // CHECK2: omp.inner.for.inc:
1570 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1571 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1572 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1573 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1574 // CHECK2: omp.inner.for.end:
1575 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1576 // CHECK2: omp.loop.exit:
1577 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1578 // CHECK2-NEXT: ret void
1579 //
1580 //
1581 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
1582 // CHECK2-SAME: () #[[ATTR1]] {
1583 // CHECK2-NEXT: entry:
1584 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
1585 // CHECK2-NEXT: ret void
1586 //
1587 //
1588 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1589 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1590 // CHECK2-NEXT: entry:
1591 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1592 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1593 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1594 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1595 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1596 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1597 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1598 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1599 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1600 // CHECK2-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1601 // CHECK2-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1602 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1603 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1604 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1605 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1606 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1607 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1608 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1609 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1610 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1611 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1612 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1613 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1614 // CHECK2: cond.true:
1615 // CHECK2-NEXT: br label [[COND_END:%.*]]
1616 // CHECK2: cond.false:
1617 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1618 // CHECK2-NEXT: br label [[COND_END]]
1619 // CHECK2: cond.end:
1620 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1621 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1622 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1623 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1624 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1625 // CHECK2: omp.inner.for.cond:
1626 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1627 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1628 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1629 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1630 // CHECK2: omp.inner.for.body:
1631 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1632 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1633 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1634 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1635 // CHECK2-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1636 // CHECK2-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1637 // CHECK2-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
1638 // CHECK2-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1639 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1640 // CHECK2: omp.inner.for.inc:
1641 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1642 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1643 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1644 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1645 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1646 // CHECK2: omp.inner.for.end:
1647 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1648 // CHECK2: omp.loop.exit:
1649 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1650 // CHECK2-NEXT: ret void
1651 //
1652 //
1653 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1654 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1655 // CHECK2-NEXT: entry:
1656 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1657 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1658 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1659 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1660 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1661 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1662 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1663 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1664 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1665 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1666 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1667 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1668 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1669 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1670 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1671 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1672 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1673 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1674 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1675 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1676 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1677 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1678 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1679 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1680 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1681 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1682 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1683 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1684 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1685 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1686 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1687 // CHECK2: cond.true:
1688 // CHECK2-NEXT: br label [[COND_END:%.*]]
1689 // CHECK2: cond.false:
1690 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1691 // CHECK2-NEXT: br label [[COND_END]]
1692 // CHECK2: cond.end:
1693 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1694 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1695 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1696 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1697 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1698 // CHECK2: omp.inner.for.cond:
1699 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1700 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1701 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1702 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1703 // CHECK2: omp.inner.for.body:
1704 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1705 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1706 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1707 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1708 // CHECK2-NEXT: call void @_Z9gtid_testv()
1709 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1710 // CHECK2: omp.body.continue:
1711 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1712 // CHECK2: omp.inner.for.inc:
1713 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1714 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1715 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1716 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1717 // CHECK2: omp.inner.for.end:
1718 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1719 // CHECK2: omp.loop.exit:
1720 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1721 // CHECK2-NEXT: ret void
1722 //
1723 //
1724 // CHECK2-LABEL: define {{[^@]+}}@main
1725 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
1726 // CHECK2-NEXT: entry:
1727 // CHECK2-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1728 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1729 // CHECK2-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
1730 // CHECK2-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
1731 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1732 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1733 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1734 // CHECK2-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
1735 // CHECK2-NEXT: store i32 0, i32* [[RETVAL]], align 4
1736 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1737 // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
1738 // CHECK2-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1739 // CHECK2-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1740 // CHECK2: omp_offload.failed:
1741 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
1742 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
1743 // CHECK2: omp_offload.cont:
1744 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1745 // CHECK2-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
1746 // CHECK2-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1747 // CHECK2-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1748 // CHECK2: omp_offload.failed2:
1749 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
1750 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT3]]
1751 // CHECK2: omp_offload.cont3:
1752 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
1753 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
1754 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
1755 // CHECK2-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
1756 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1757 // CHECK2-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1758 // CHECK2-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
1759 // CHECK2-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1760 // CHECK2-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1761 // CHECK2-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
1762 // CHECK2-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1763 // CHECK2-NEXT: store i8* null, i8** [[TMP10]], align 8
1764 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1765 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1766 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1767 // CHECK2-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1768 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1769 // CHECK2-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1770 // CHECK2: omp_offload.failed5:
1771 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
1772 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT6]]
1773 // CHECK2: omp_offload.cont6:
1774 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
1775 // CHECK2-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
1776 // CHECK2-NEXT: ret i32 [[CALL]]
1777 //
1778 //
1779 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
1780 // CHECK2-SAME: () #[[ATTR1]] {
1781 // CHECK2-NEXT: entry:
1782 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1783 // CHECK2-NEXT: ret void
1784 //
1785 //
1786 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1787 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1788 // CHECK2-NEXT: entry:
1789 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1790 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1791 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1792 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1793 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1794 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1795 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1796 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1797 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1798 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1799 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1800 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1801 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1802 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1803 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1804 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1805 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1806 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1807 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1808 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1809 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1810 // CHECK2: cond.true:
1811 // CHECK2-NEXT: br label [[COND_END:%.*]]
1812 // CHECK2: cond.false:
1813 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1814 // CHECK2-NEXT: br label [[COND_END]]
1815 // CHECK2: cond.end:
1816 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1817 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1818 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1819 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1820 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1821 // CHECK2: omp.inner.for.cond:
1822 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1823 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1824 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1825 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1826 // CHECK2: omp.inner.for.body:
1827 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1828 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1829 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1830 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1831 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
1832 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1833 // CHECK2: omp.inner.for.inc:
1834 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1835 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1836 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1837 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1838 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1839 // CHECK2: omp.inner.for.end:
1840 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1841 // CHECK2: omp.loop.exit:
1842 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1843 // CHECK2-NEXT: ret void
1844 //
1845 //
1846 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
1847 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1848 // CHECK2-NEXT: entry:
1849 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1850 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1851 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1852 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1853 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1854 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1855 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1856 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1857 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1858 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1859 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1860 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1861 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1862 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1863 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1864 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1865 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
1866 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1867 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1868 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1869 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1870 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1871 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1872 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1873 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1874 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1875 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1876 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1877 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1878 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1879 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1880 // CHECK2: cond.true:
1881 // CHECK2-NEXT: br label [[COND_END:%.*]]
1882 // CHECK2: cond.false:
1883 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1884 // CHECK2-NEXT: br label [[COND_END]]
1885 // CHECK2: cond.end:
1886 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1887 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1888 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1889 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1890 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1891 // CHECK2: omp.inner.for.cond:
1892 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1893 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1894 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1895 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1896 // CHECK2: omp.inner.for.body:
1897 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1898 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1899 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1900 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1901 // CHECK2-NEXT: call void @_Z3fn4v()
1902 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1903 // CHECK2: omp.body.continue:
1904 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1905 // CHECK2: omp.inner.for.inc:
1906 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1907 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1908 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1909 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1910 // CHECK2: omp.inner.for.end:
1911 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1912 // CHECK2: omp.loop.exit:
1913 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1914 // CHECK2-NEXT: ret void
1915 //
1916 //
1917 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
1918 // CHECK2-SAME: () #[[ATTR1]] {
1919 // CHECK2-NEXT: entry:
1920 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
1921 // CHECK2-NEXT: ret void
1922 //
1923 //
1924 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
1925 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1926 // CHECK2-NEXT: entry:
1927 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1928 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1929 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1930 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1931 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1932 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1933 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1934 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1935 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1936 // CHECK2-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1937 // CHECK2-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1938 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1939 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1940 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1941 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1942 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1943 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1944 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1945 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1946 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1947 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1948 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1949 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1950 // CHECK2: cond.true:
1951 // CHECK2-NEXT: br label [[COND_END:%.*]]
1952 // CHECK2: cond.false:
1953 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1954 // CHECK2-NEXT: br label [[COND_END]]
1955 // CHECK2: cond.end:
1956 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1957 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1958 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1959 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1960 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1961 // CHECK2: omp.inner.for.cond:
1962 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1963 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1964 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1965 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1966 // CHECK2: omp.inner.for.body:
1967 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1968 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1969 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1970 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1971 // CHECK2-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1972 // CHECK2-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1973 // CHECK2-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
1974 // CHECK2-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1975 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1976 // CHECK2: omp.inner.for.inc:
1977 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1978 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1979 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1980 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1981 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1982 // CHECK2: omp.inner.for.end:
1983 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1984 // CHECK2: omp.loop.exit:
1985 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1986 // CHECK2-NEXT: ret void
1987 //
1988 //
1989 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
1990 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1991 // CHECK2-NEXT: entry:
1992 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1993 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1994 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1995 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1996 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1997 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1998 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1999 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2000 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2001 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2002 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2003 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2004 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2005 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2006 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2007 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2008 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2009 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2010 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2011 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2012 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2013 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2014 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2015 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2016 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2017 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2018 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2019 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2020 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2021 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2022 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2023 // CHECK2: cond.true:
2024 // CHECK2-NEXT: br label [[COND_END:%.*]]
2025 // CHECK2: cond.false:
2026 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2027 // CHECK2-NEXT: br label [[COND_END]]
2028 // CHECK2: cond.end:
2029 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2030 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2031 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2032 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2033 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2034 // CHECK2: omp.inner.for.cond:
2035 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2036 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2037 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2038 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2039 // CHECK2: omp.inner.for.body:
2040 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2041 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2042 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2043 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2044 // CHECK2-NEXT: call void @_Z3fn5v()
2045 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2046 // CHECK2: omp.body.continue:
2047 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2048 // CHECK2: omp.inner.for.inc:
2049 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2050 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2051 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2052 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2053 // CHECK2: omp.inner.for.end:
2054 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2055 // CHECK2: omp.loop.exit:
2056 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2057 // CHECK2-NEXT: ret void
2058 //
2059 //
2060 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
2061 // CHECK2-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
2062 // CHECK2-NEXT: entry:
2063 // CHECK2-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
2064 // CHECK2-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
2065 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
2066 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
2067 // CHECK2-NEXT: ret void
2068 //
2069 //
2070 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..8
2071 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
2072 // CHECK2-NEXT: entry:
2073 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2074 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2075 // CHECK2-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
2076 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2077 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2078 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2079 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2080 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2081 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2082 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2083 // CHECK2-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2084 // CHECK2-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2085 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2086 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2087 // CHECK2-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
2088 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
2089 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2090 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2091 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2092 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2093 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2094 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2095 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2096 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2097 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
2098 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2099 // CHECK2: cond.true:
2100 // CHECK2-NEXT: br label [[COND_END:%.*]]
2101 // CHECK2: cond.false:
2102 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2103 // CHECK2-NEXT: br label [[COND_END]]
2104 // CHECK2: cond.end:
2105 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2106 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2107 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2108 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2109 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2110 // CHECK2: omp.inner.for.cond:
2111 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2112 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2113 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2114 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2115 // CHECK2: omp.inner.for.body:
2116 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2117 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2118 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2119 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2120 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
2121 // CHECK2-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
2122 // CHECK2-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2123 // CHECK2: omp_if.then:
2124 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
2125 // CHECK2-NEXT: br label [[OMP_IF_END:%.*]]
2126 // CHECK2: omp_if.else:
2127 // CHECK2-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
2128 // CHECK2-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2129 // CHECK2-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
2130 // CHECK2-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
2131 // CHECK2-NEXT: br label [[OMP_IF_END]]
2132 // CHECK2: omp_if.end:
2133 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2134 // CHECK2: omp.inner.for.inc:
2135 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2136 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2137 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2138 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2139 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2140 // CHECK2: omp.inner.for.end:
2141 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2142 // CHECK2: omp.loop.exit:
2143 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2144 // CHECK2-NEXT: ret void
2145 //
2146 //
2147 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
2148 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2149 // CHECK2-NEXT: entry:
2150 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2151 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2152 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2153 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2154 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2155 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2156 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2157 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2158 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2159 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2160 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2161 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2162 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2163 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2164 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2165 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2166 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2167 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2168 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2169 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2170 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2171 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2172 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2173 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2174 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2175 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2176 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2177 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2178 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2179 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2180 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2181 // CHECK2: cond.true:
2182 // CHECK2-NEXT: br label [[COND_END:%.*]]
2183 // CHECK2: cond.false:
2184 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2185 // CHECK2-NEXT: br label [[COND_END]]
2186 // CHECK2: cond.end:
2187 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2188 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2189 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2190 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2191 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2192 // CHECK2: omp.inner.for.cond:
2193 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2194 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2195 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2196 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2197 // CHECK2: omp.inner.for.body:
2198 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2199 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2200 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2201 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2202 // CHECK2-NEXT: call void @_Z3fn6v()
2203 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2204 // CHECK2: omp.body.continue:
2205 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2206 // CHECK2: omp.inner.for.inc:
2207 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2208 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2209 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2210 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2211 // CHECK2: omp.inner.for.end:
2212 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2213 // CHECK2: omp.loop.exit:
2214 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2215 // CHECK2-NEXT: ret void
2216 //
2217 //
2218 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
2219 // CHECK2-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
2220 // CHECK2-NEXT: entry:
2221 // CHECK2-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
2222 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2223 // CHECK2-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
2224 // CHECK2-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
2225 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2226 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2227 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2228 // CHECK2-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
2229 // CHECK2-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
2230 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2231 // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
2232 // CHECK2-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
2233 // CHECK2-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2234 // CHECK2: omp_offload.failed:
2235 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
2236 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
2237 // CHECK2: omp_offload.cont:
2238 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2239 // CHECK2-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
2240 // CHECK2-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
2241 // CHECK2-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
2242 // CHECK2: omp_offload.failed2:
2243 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
2244 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT3]]
2245 // CHECK2: omp_offload.cont3:
2246 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
2247 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
2248 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
2249 // CHECK2-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
2250 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2251 // CHECK2-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
2252 // CHECK2-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
2253 // CHECK2-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2254 // CHECK2-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
2255 // CHECK2-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
2256 // CHECK2-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2257 // CHECK2-NEXT: store i8* null, i8** [[TMP10]], align 8
2258 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2259 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2260 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2261 // CHECK2-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2262 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2263 // CHECK2-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
2264 // CHECK2: omp_offload.failed5:
2265 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
2266 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT6]]
2267 // CHECK2: omp_offload.cont6:
2268 // CHECK2-NEXT: ret i32 0
2269 //
2270 //
2271 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
2272 // CHECK2-SAME: () #[[ATTR1]] {
2273 // CHECK2-NEXT: entry:
2274 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
2275 // CHECK2-NEXT: ret void
2276 //
2277 //
2278 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
2279 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2280 // CHECK2-NEXT: entry:
2281 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2282 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2283 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2284 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2285 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2286 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2287 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2288 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2289 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2290 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2291 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2292 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2293 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2294 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2295 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2296 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2297 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2298 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2299 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2300 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2301 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2302 // CHECK2: cond.true:
2303 // CHECK2-NEXT: br label [[COND_END:%.*]]
2304 // CHECK2: cond.false:
2305 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2306 // CHECK2-NEXT: br label [[COND_END]]
2307 // CHECK2: cond.end:
2308 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2309 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2310 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2311 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2312 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2313 // CHECK2: omp.inner.for.cond:
2314 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2315 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2316 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2317 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2318 // CHECK2: omp.inner.for.body:
2319 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2320 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2321 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2322 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2323 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
2324 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2325 // CHECK2: omp.inner.for.inc:
2326 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2327 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2328 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2329 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2330 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2331 // CHECK2: omp.inner.for.end:
2332 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2333 // CHECK2: omp.loop.exit:
2334 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2335 // CHECK2-NEXT: ret void
2336 //
2337 //
2338 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
2339 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2340 // CHECK2-NEXT: entry:
2341 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2342 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2343 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2344 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2345 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2346 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2347 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2348 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2349 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2350 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2351 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2352 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2353 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2354 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2355 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2356 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2357 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2358 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2359 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2360 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2361 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2362 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2363 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2364 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2365 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2366 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2367 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2368 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2369 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2370 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2371 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2372 // CHECK2: cond.true:
2373 // CHECK2-NEXT: br label [[COND_END:%.*]]
2374 // CHECK2: cond.false:
2375 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2376 // CHECK2-NEXT: br label [[COND_END]]
2377 // CHECK2: cond.end:
2378 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2379 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2380 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2381 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2382 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2383 // CHECK2: omp.inner.for.cond:
2384 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2385 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2386 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2387 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2388 // CHECK2: omp.inner.for.body:
2389 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2390 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2391 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2392 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2393 // CHECK2-NEXT: call void @_Z3fn1v()
2394 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2395 // CHECK2: omp.body.continue:
2396 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2397 // CHECK2: omp.inner.for.inc:
2398 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2399 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2400 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2401 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2402 // CHECK2: omp.inner.for.end:
2403 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2404 // CHECK2: omp.loop.exit:
2405 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2406 // CHECK2-NEXT: ret void
2407 //
2408 //
2409 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
2410 // CHECK2-SAME: () #[[ATTR1]] {
2411 // CHECK2-NEXT: entry:
2412 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
2413 // CHECK2-NEXT: ret void
2414 //
2415 //
2416 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..12
2417 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2418 // CHECK2-NEXT: entry:
2419 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2420 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2421 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2422 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2423 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2424 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2425 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2426 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2427 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2428 // CHECK2-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2429 // CHECK2-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2430 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2431 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2432 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2433 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2434 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2435 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2436 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2437 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2438 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2439 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2440 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2441 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2442 // CHECK2: cond.true:
2443 // CHECK2-NEXT: br label [[COND_END:%.*]]
2444 // CHECK2: cond.false:
2445 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2446 // CHECK2-NEXT: br label [[COND_END]]
2447 // CHECK2: cond.end:
2448 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2449 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2450 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2451 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2452 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2453 // CHECK2: omp.inner.for.cond:
2454 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2455 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2456 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2457 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2458 // CHECK2: omp.inner.for.body:
2459 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2460 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2461 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2462 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2463 // CHECK2-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2464 // CHECK2-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2465 // CHECK2-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
2466 // CHECK2-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2467 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2468 // CHECK2: omp.inner.for.inc:
2469 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2470 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2471 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2472 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2473 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2474 // CHECK2: omp.inner.for.end:
2475 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2476 // CHECK2: omp.loop.exit:
2477 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2478 // CHECK2-NEXT: ret void
2479 //
2480 //
2481 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..13
2482 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2483 // CHECK2-NEXT: entry:
2484 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2485 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2486 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2487 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2488 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2489 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2490 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2491 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2492 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2493 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2494 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2495 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2496 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2497 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2498 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2499 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2500 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2501 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2502 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2503 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2504 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2505 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2506 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2507 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2508 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2509 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2510 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2511 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2512 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2513 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2514 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2515 // CHECK2: cond.true:
2516 // CHECK2-NEXT: br label [[COND_END:%.*]]
2517 // CHECK2: cond.false:
2518 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2519 // CHECK2-NEXT: br label [[COND_END]]
2520 // CHECK2: cond.end:
2521 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2522 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2523 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2524 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2525 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2526 // CHECK2: omp.inner.for.cond:
2527 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2528 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2529 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2530 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2531 // CHECK2: omp.inner.for.body:
2532 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2533 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2534 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2535 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2536 // CHECK2-NEXT: call void @_Z3fn2v()
2537 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2538 // CHECK2: omp.body.continue:
2539 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2540 // CHECK2: omp.inner.for.inc:
2541 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2542 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2543 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2544 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2545 // CHECK2: omp.inner.for.end:
2546 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2547 // CHECK2: omp.loop.exit:
2548 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2549 // CHECK2-NEXT: ret void
2550 //
2551 //
2552 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
2553 // CHECK2-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
2554 // CHECK2-NEXT: entry:
2555 // CHECK2-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
2556 // CHECK2-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
2557 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
2558 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
2559 // CHECK2-NEXT: ret void
2560 //
2561 //
2562 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
2563 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
2564 // CHECK2-NEXT: entry:
2565 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2566 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2567 // CHECK2-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
2568 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2569 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2570 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2571 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2572 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2573 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2574 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2575 // CHECK2-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2576 // CHECK2-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2577 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2578 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2579 // CHECK2-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
2580 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
2581 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2582 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2583 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2584 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2585 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2586 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2587 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2588 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2589 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
2590 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2591 // CHECK2: cond.true:
2592 // CHECK2-NEXT: br label [[COND_END:%.*]]
2593 // CHECK2: cond.false:
2594 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2595 // CHECK2-NEXT: br label [[COND_END]]
2596 // CHECK2: cond.end:
2597 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2598 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2599 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2600 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2601 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2602 // CHECK2: omp.inner.for.cond:
2603 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2604 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2605 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2606 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2607 // CHECK2: omp.inner.for.body:
2608 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2609 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2610 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2611 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2612 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
2613 // CHECK2-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
2614 // CHECK2-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2615 // CHECK2: omp_if.then:
2616 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
2617 // CHECK2-NEXT: br label [[OMP_IF_END:%.*]]
2618 // CHECK2: omp_if.else:
2619 // CHECK2-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
2620 // CHECK2-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2621 // CHECK2-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
2622 // CHECK2-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
2623 // CHECK2-NEXT: br label [[OMP_IF_END]]
2624 // CHECK2: omp_if.end:
2625 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2626 // CHECK2: omp.inner.for.inc:
2627 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2628 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2629 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2630 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2631 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2632 // CHECK2: omp.inner.for.end:
2633 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2634 // CHECK2: omp.loop.exit:
2635 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2636 // CHECK2-NEXT: ret void
2637 //
2638 //
2639 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
2640 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2641 // CHECK2-NEXT: entry:
2642 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2643 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2644 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2645 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2646 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2647 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
2648 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2649 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2650 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2651 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2652 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
2653 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2654 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2655 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2656 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2657 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2658 // CHECK2-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2659 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2660 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2661 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2662 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2663 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2664 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2665 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2666 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2667 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2668 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2669 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2670 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2671 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2672 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2673 // CHECK2: cond.true:
2674 // CHECK2-NEXT: br label [[COND_END:%.*]]
2675 // CHECK2: cond.false:
2676 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2677 // CHECK2-NEXT: br label [[COND_END]]
2678 // CHECK2: cond.end:
2679 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2680 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2681 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2682 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2683 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2684 // CHECK2: omp.inner.for.cond:
2685 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2686 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2687 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2688 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2689 // CHECK2: omp.inner.for.body:
2690 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2691 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2692 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2693 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2694 // CHECK2-NEXT: call void @_Z3fn3v()
2695 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2696 // CHECK2: omp.body.continue:
2697 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2698 // CHECK2: omp.inner.for.inc:
2699 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2700 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2701 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2702 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
2703 // CHECK2: omp.inner.for.end:
2704 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2705 // CHECK2: omp.loop.exit:
2706 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2707 // CHECK2-NEXT: ret void
2708 //
2709 //
2710 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2711 // CHECK2-SAME: () #[[ATTR5:[0-9]+]] {
2712 // CHECK2-NEXT: entry:
2713 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1)
2714 // CHECK2-NEXT: ret void
2715 //
2716 //
2717 // CHECK5-LABEL: define {{[^@]+}}@_Z9gtid_testv
2718 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
2719 // CHECK5-NEXT: entry:
2720 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2721 // CHECK5-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
2722 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
2723 // CHECK5-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
2724 // CHECK5-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
2725 // CHECK5-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2726 // CHECK5: omp_offload.failed:
2727 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
2728 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT]]
2729 // CHECK5: omp_offload.cont:
2730 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2731 // CHECK5-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
2732 // CHECK5-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
2733 // CHECK5-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
2734 // CHECK5: omp_offload.failed2:
2735 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
2736 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT3]]
2737 // CHECK5: omp_offload.cont3:
2738 // CHECK5-NEXT: ret void
2739 //
2740 //
2741 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
2742 // CHECK5-SAME: () #[[ATTR1:[0-9]+]] {
2743 // CHECK5-NEXT: entry:
2744 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
2745 // CHECK5-NEXT: ret void
2746 //
2747 //
2748 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
2749 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2750 // CHECK5-NEXT: entry:
2751 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2752 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2753 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2754 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2755 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2756 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2757 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2758 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2759 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
2760 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2761 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2762 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2763 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2764 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2765 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2766 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2767 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2768 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2769 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2770 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2771 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2772 // CHECK5: cond.true:
2773 // CHECK5-NEXT: br label [[COND_END:%.*]]
2774 // CHECK5: cond.false:
2775 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2776 // CHECK5-NEXT: br label [[COND_END]]
2777 // CHECK5: cond.end:
2778 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2779 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2780 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2781 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2782 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2783 // CHECK5: omp.inner.for.cond:
2784 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2785 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2786 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2787 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2788 // CHECK5: omp.inner.for.body:
2789 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2790 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2791 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2792 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2793 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
2794 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2795 // CHECK5: omp.inner.for.inc:
2796 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2797 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2798 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2799 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2800 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
2801 // CHECK5: omp.inner.for.end:
2802 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2803 // CHECK5: omp.loop.exit:
2804 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2805 // CHECK5-NEXT: ret void
2806 //
2807 //
2808 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
2809 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2810 // CHECK5-NEXT: entry:
2811 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2812 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2813 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2814 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2815 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2816 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2817 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2818 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2819 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2820 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2821 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
2822 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2823 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2824 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2825 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2826 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2827 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2828 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2829 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2830 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2831 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2832 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2833 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2834 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2835 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2836 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2837 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2838 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2839 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2840 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2841 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2842 // CHECK5: cond.true:
2843 // CHECK5-NEXT: br label [[COND_END:%.*]]
2844 // CHECK5: cond.false:
2845 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2846 // CHECK5-NEXT: br label [[COND_END]]
2847 // CHECK5: cond.end:
2848 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2849 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2850 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2851 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2852 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2853 // CHECK5: omp.inner.for.cond:
2854 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2855 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2856 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2857 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2858 // CHECK5: omp.inner.for.body:
2859 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2860 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2861 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2862 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2863 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2864 // CHECK5: omp.body.continue:
2865 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2866 // CHECK5: omp.inner.for.inc:
2867 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2868 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2869 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2870 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
2871 // CHECK5: omp.inner.for.end:
2872 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2873 // CHECK5: omp.loop.exit:
2874 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2875 // CHECK5-NEXT: ret void
2876 //
2877 //
2878 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
2879 // CHECK5-SAME: () #[[ATTR1]] {
2880 // CHECK5-NEXT: entry:
2881 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
2882 // CHECK5-NEXT: ret void
2883 //
2884 //
2885 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
2886 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2887 // CHECK5-NEXT: entry:
2888 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2889 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2890 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2891 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2892 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2893 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2894 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2895 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2896 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
2897 // CHECK5-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2898 // CHECK5-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2899 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2900 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2901 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2902 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2903 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2904 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2905 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2906 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2907 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2908 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2909 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2910 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2911 // CHECK5: cond.true:
2912 // CHECK5-NEXT: br label [[COND_END:%.*]]
2913 // CHECK5: cond.false:
2914 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2915 // CHECK5-NEXT: br label [[COND_END]]
2916 // CHECK5: cond.end:
2917 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2918 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2919 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2920 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2921 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2922 // CHECK5: omp.inner.for.cond:
2923 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2924 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2925 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2926 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2927 // CHECK5: omp.inner.for.body:
2928 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2929 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2930 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2931 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2932 // CHECK5-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2933 // CHECK5-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2934 // CHECK5-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
2935 // CHECK5-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2936 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2937 // CHECK5: omp.inner.for.inc:
2938 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2939 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2940 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2941 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2942 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
2943 // CHECK5: omp.inner.for.end:
2944 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2945 // CHECK5: omp.loop.exit:
2946 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2947 // CHECK5-NEXT: ret void
2948 //
2949 //
2950 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
2951 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2952 // CHECK5-NEXT: entry:
2953 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2954 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2955 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2956 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2957 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2958 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2959 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2960 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2961 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2962 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2963 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
2964 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2965 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2966 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2967 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2968 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2969 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
2970 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2971 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2972 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2973 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2974 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2975 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2976 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2977 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2978 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2979 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2980 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2981 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2982 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2983 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2984 // CHECK5: cond.true:
2985 // CHECK5-NEXT: br label [[COND_END:%.*]]
2986 // CHECK5: cond.false:
2987 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2988 // CHECK5-NEXT: br label [[COND_END]]
2989 // CHECK5: cond.end:
2990 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2991 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2992 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2993 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2994 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2995 // CHECK5: omp.inner.for.cond:
2996 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2997 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2998 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2999 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3000 // CHECK5: omp.inner.for.body:
3001 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3002 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3003 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3004 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3005 // CHECK5-NEXT: call void @_Z9gtid_testv()
3006 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3007 // CHECK5: omp.body.continue:
3008 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3009 // CHECK5: omp.inner.for.inc:
3010 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3011 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3012 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3013 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3014 // CHECK5: omp.inner.for.end:
3015 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3016 // CHECK5: omp.loop.exit:
3017 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3018 // CHECK5-NEXT: ret void
3019 //
3020 //
3021 // CHECK5-LABEL: define {{[^@]+}}@main
3022 // CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
3023 // CHECK5-NEXT: entry:
3024 // CHECK5-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3025 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3026 // CHECK5-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
3027 // CHECK5-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
3028 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3029 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3030 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3031 // CHECK5-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
3032 // CHECK5-NEXT: store i32 0, i32* [[RETVAL]], align 4
3033 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3034 // CHECK5-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
3035 // CHECK5-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3036 // CHECK5-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3037 // CHECK5: omp_offload.failed:
3038 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
3039 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT]]
3040 // CHECK5: omp_offload.cont:
3041 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3042 // CHECK5-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
3043 // CHECK5-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
3044 // CHECK5-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
3045 // CHECK5: omp_offload.failed2:
3046 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
3047 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT3]]
3048 // CHECK5: omp_offload.cont3:
3049 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
3050 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
3051 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
3052 // CHECK5-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
3053 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3054 // CHECK5-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3055 // CHECK5-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
3056 // CHECK5-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3057 // CHECK5-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
3058 // CHECK5-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
3059 // CHECK5-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3060 // CHECK5-NEXT: store i8* null, i8** [[TMP10]], align 8
3061 // CHECK5-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3062 // CHECK5-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3063 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3064 // CHECK5-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3065 // CHECK5-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3066 // CHECK5-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
3067 // CHECK5: omp_offload.failed5:
3068 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
3069 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT6]]
3070 // CHECK5: omp_offload.cont6:
3071 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
3072 // CHECK5-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
3073 // CHECK5-NEXT: ret i32 [[CALL]]
3074 //
3075 //
3076 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
3077 // CHECK5-SAME: () #[[ATTR1]] {
3078 // CHECK5-NEXT: entry:
3079 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
3080 // CHECK5-NEXT: ret void
3081 //
3082 //
3083 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..4
3084 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3085 // CHECK5-NEXT: entry:
3086 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3087 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3088 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3089 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3090 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3091 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3092 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3093 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3094 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3095 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3096 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3097 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3098 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3099 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3100 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3101 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3102 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3103 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3104 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3105 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3106 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3107 // CHECK5: cond.true:
3108 // CHECK5-NEXT: br label [[COND_END:%.*]]
3109 // CHECK5: cond.false:
3110 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3111 // CHECK5-NEXT: br label [[COND_END]]
3112 // CHECK5: cond.end:
3113 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3114 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3115 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3116 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3117 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3118 // CHECK5: omp.inner.for.cond:
3119 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3120 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3121 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3122 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3123 // CHECK5: omp.inner.for.body:
3124 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3125 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3126 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3127 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3128 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
3129 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3130 // CHECK5: omp.inner.for.inc:
3131 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3132 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3133 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
3134 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3135 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3136 // CHECK5: omp.inner.for.end:
3137 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3138 // CHECK5: omp.loop.exit:
3139 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3140 // CHECK5-NEXT: ret void
3141 //
3142 //
3143 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..5
3144 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3145 // CHECK5-NEXT: entry:
3146 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3147 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3148 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3149 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3150 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3151 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3152 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3153 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3154 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3155 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3156 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3157 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3158 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3159 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3160 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3161 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3162 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3163 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3164 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3165 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3166 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3167 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3168 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3169 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3170 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3171 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3172 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3173 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3174 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3175 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3176 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3177 // CHECK5: cond.true:
3178 // CHECK5-NEXT: br label [[COND_END:%.*]]
3179 // CHECK5: cond.false:
3180 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3181 // CHECK5-NEXT: br label [[COND_END]]
3182 // CHECK5: cond.end:
3183 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3184 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3185 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3186 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3187 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3188 // CHECK5: omp.inner.for.cond:
3189 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3190 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3191 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3192 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3193 // CHECK5: omp.inner.for.body:
3194 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3195 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3196 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3197 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3198 // CHECK5-NEXT: call void @_Z3fn4v()
3199 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3200 // CHECK5: omp.body.continue:
3201 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3202 // CHECK5: omp.inner.for.inc:
3203 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3204 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3205 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3206 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3207 // CHECK5: omp.inner.for.end:
3208 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3209 // CHECK5: omp.loop.exit:
3210 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3211 // CHECK5-NEXT: ret void
3212 //
3213 //
3214 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
3215 // CHECK5-SAME: () #[[ATTR1]] {
3216 // CHECK5-NEXT: entry:
3217 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
3218 // CHECK5-NEXT: ret void
3219 //
3220 //
3221 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
3222 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3223 // CHECK5-NEXT: entry:
3224 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3225 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3226 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3227 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3228 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3229 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3230 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3231 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3232 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3233 // CHECK5-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3234 // CHECK5-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
3235 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3236 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3237 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3238 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3239 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3240 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3241 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3242 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3243 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3244 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3245 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3246 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3247 // CHECK5: cond.true:
3248 // CHECK5-NEXT: br label [[COND_END:%.*]]
3249 // CHECK5: cond.false:
3250 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3251 // CHECK5-NEXT: br label [[COND_END]]
3252 // CHECK5: cond.end:
3253 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3254 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3255 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3256 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3257 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3258 // CHECK5: omp.inner.for.cond:
3259 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3260 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3261 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3262 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3263 // CHECK5: omp.inner.for.body:
3264 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3265 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3266 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3267 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3268 // CHECK5-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3269 // CHECK5-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3270 // CHECK5-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
3271 // CHECK5-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3272 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3273 // CHECK5: omp.inner.for.inc:
3274 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3275 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3276 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3277 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3278 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3279 // CHECK5: omp.inner.for.end:
3280 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3281 // CHECK5: omp.loop.exit:
3282 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3283 // CHECK5-NEXT: ret void
3284 //
3285 //
3286 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
3287 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3288 // CHECK5-NEXT: entry:
3289 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3290 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3291 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3292 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3293 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3294 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3295 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3296 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3297 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3298 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3299 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3300 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3301 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3302 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3303 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3304 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3305 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3306 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3307 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3308 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3309 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3310 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3311 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3312 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3313 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3314 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3315 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3316 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3317 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3318 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3319 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3320 // CHECK5: cond.true:
3321 // CHECK5-NEXT: br label [[COND_END:%.*]]
3322 // CHECK5: cond.false:
3323 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3324 // CHECK5-NEXT: br label [[COND_END]]
3325 // CHECK5: cond.end:
3326 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3327 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3328 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3329 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3330 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3331 // CHECK5: omp.inner.for.cond:
3332 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3333 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3334 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3335 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3336 // CHECK5: omp.inner.for.body:
3337 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3338 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3339 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3340 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3341 // CHECK5-NEXT: call void @_Z3fn5v()
3342 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3343 // CHECK5: omp.body.continue:
3344 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3345 // CHECK5: omp.inner.for.inc:
3346 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3347 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3348 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3349 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3350 // CHECK5: omp.inner.for.end:
3351 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3352 // CHECK5: omp.loop.exit:
3353 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3354 // CHECK5-NEXT: ret void
3355 //
3356 //
3357 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
3358 // CHECK5-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
3359 // CHECK5-NEXT: entry:
3360 // CHECK5-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
3361 // CHECK5-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
3362 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
3363 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
3364 // CHECK5-NEXT: ret void
3365 //
3366 //
3367 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..8
3368 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
3369 // CHECK5-NEXT: entry:
3370 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3371 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3372 // CHECK5-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
3373 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3374 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3375 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3376 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3377 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3378 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3379 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3380 // CHECK5-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3381 // CHECK5-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
3382 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3383 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3384 // CHECK5-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
3385 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
3386 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3387 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3388 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3389 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3390 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3391 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3392 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3393 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3394 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
3395 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3396 // CHECK5: cond.true:
3397 // CHECK5-NEXT: br label [[COND_END:%.*]]
3398 // CHECK5: cond.false:
3399 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3400 // CHECK5-NEXT: br label [[COND_END]]
3401 // CHECK5: cond.end:
3402 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3403 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3404 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3405 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3406 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3407 // CHECK5: omp.inner.for.cond:
3408 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3409 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3410 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3411 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3412 // CHECK5: omp.inner.for.body:
3413 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3414 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
3415 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3416 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3417 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
3418 // CHECK5-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
3419 // CHECK5-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3420 // CHECK5: omp_if.then:
3421 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
3422 // CHECK5-NEXT: br label [[OMP_IF_END:%.*]]
3423 // CHECK5: omp_if.else:
3424 // CHECK5-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
3425 // CHECK5-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3426 // CHECK5-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
3427 // CHECK5-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
3428 // CHECK5-NEXT: br label [[OMP_IF_END]]
3429 // CHECK5: omp_if.end:
3430 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3431 // CHECK5: omp.inner.for.inc:
3432 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3433 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3434 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
3435 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3436 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3437 // CHECK5: omp.inner.for.end:
3438 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3439 // CHECK5: omp.loop.exit:
3440 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3441 // CHECK5-NEXT: ret void
3442 //
3443 //
3444 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..9
3445 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3446 // CHECK5-NEXT: entry:
3447 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3448 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3449 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3450 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3451 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3452 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3453 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3454 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3455 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3456 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3457 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3458 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3459 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3460 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3461 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3462 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3463 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3464 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3465 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3466 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3467 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3468 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3469 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3470 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3471 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3472 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3473 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3474 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3475 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3476 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3477 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3478 // CHECK5: cond.true:
3479 // CHECK5-NEXT: br label [[COND_END:%.*]]
3480 // CHECK5: cond.false:
3481 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3482 // CHECK5-NEXT: br label [[COND_END]]
3483 // CHECK5: cond.end:
3484 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3485 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3486 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3487 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3488 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3489 // CHECK5: omp.inner.for.cond:
3490 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3491 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3492 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3493 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3494 // CHECK5: omp.inner.for.body:
3495 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3496 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3497 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3498 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3499 // CHECK5-NEXT: call void @_Z3fn6v()
3500 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3501 // CHECK5: omp.body.continue:
3502 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3503 // CHECK5: omp.inner.for.inc:
3504 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3505 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3506 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3507 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3508 // CHECK5: omp.inner.for.end:
3509 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3510 // CHECK5: omp.loop.exit:
3511 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3512 // CHECK5-NEXT: ret void
3513 //
3514 //
3515 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
3516 // CHECK5-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
3517 // CHECK5-NEXT: entry:
3518 // CHECK5-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
3519 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3520 // CHECK5-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
3521 // CHECK5-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
3522 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3523 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3524 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3525 // CHECK5-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
3526 // CHECK5-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
3527 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3528 // CHECK5-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
3529 // CHECK5-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3530 // CHECK5-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3531 // CHECK5: omp_offload.failed:
3532 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
3533 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT]]
3534 // CHECK5: omp_offload.cont:
3535 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3536 // CHECK5-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
3537 // CHECK5-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
3538 // CHECK5-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
3539 // CHECK5: omp_offload.failed2:
3540 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
3541 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT3]]
3542 // CHECK5: omp_offload.cont3:
3543 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
3544 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
3545 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
3546 // CHECK5-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
3547 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3548 // CHECK5-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3549 // CHECK5-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
3550 // CHECK5-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3551 // CHECK5-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
3552 // CHECK5-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
3553 // CHECK5-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3554 // CHECK5-NEXT: store i8* null, i8** [[TMP10]], align 8
3555 // CHECK5-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3556 // CHECK5-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3557 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3558 // CHECK5-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3559 // CHECK5-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3560 // CHECK5-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
3561 // CHECK5: omp_offload.failed5:
3562 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
3563 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT6]]
3564 // CHECK5: omp_offload.cont6:
3565 // CHECK5-NEXT: ret i32 0
3566 //
3567 //
3568 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
3569 // CHECK5-SAME: () #[[ATTR1]] {
3570 // CHECK5-NEXT: entry:
3571 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
3572 // CHECK5-NEXT: ret void
3573 //
3574 //
3575 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
3576 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3577 // CHECK5-NEXT: entry:
3578 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3579 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3580 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3581 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3582 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3583 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3584 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3585 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3586 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3587 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3588 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3589 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3590 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3591 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3592 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3593 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3594 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3595 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3596 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3597 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3598 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3599 // CHECK5: cond.true:
3600 // CHECK5-NEXT: br label [[COND_END:%.*]]
3601 // CHECK5: cond.false:
3602 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3603 // CHECK5-NEXT: br label [[COND_END]]
3604 // CHECK5: cond.end:
3605 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3606 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3607 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3608 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3609 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3610 // CHECK5: omp.inner.for.cond:
3611 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3612 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3613 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3614 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3615 // CHECK5: omp.inner.for.body:
3616 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3617 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3618 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3619 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3620 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
3621 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3622 // CHECK5: omp.inner.for.inc:
3623 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3624 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3625 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
3626 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3627 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3628 // CHECK5: omp.inner.for.end:
3629 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3630 // CHECK5: omp.loop.exit:
3631 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3632 // CHECK5-NEXT: ret void
3633 //
3634 //
3635 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..11
3636 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3637 // CHECK5-NEXT: entry:
3638 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3639 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3640 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3641 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3642 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3643 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3644 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3645 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3646 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3647 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3648 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3649 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3650 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3651 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3652 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3653 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3654 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3655 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3656 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3657 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3658 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3659 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3660 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3661 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3662 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3663 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3664 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3665 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3666 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3667 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3668 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3669 // CHECK5: cond.true:
3670 // CHECK5-NEXT: br label [[COND_END:%.*]]
3671 // CHECK5: cond.false:
3672 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3673 // CHECK5-NEXT: br label [[COND_END]]
3674 // CHECK5: cond.end:
3675 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3676 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3677 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3678 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3679 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3680 // CHECK5: omp.inner.for.cond:
3681 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3682 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3683 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3684 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3685 // CHECK5: omp.inner.for.body:
3686 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3687 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3688 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3689 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3690 // CHECK5-NEXT: call void @_Z3fn1v()
3691 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3692 // CHECK5: omp.body.continue:
3693 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3694 // CHECK5: omp.inner.for.inc:
3695 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3696 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3697 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3698 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3699 // CHECK5: omp.inner.for.end:
3700 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3701 // CHECK5: omp.loop.exit:
3702 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3703 // CHECK5-NEXT: ret void
3704 //
3705 //
3706 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
3707 // CHECK5-SAME: () #[[ATTR1]] {
3708 // CHECK5-NEXT: entry:
3709 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
3710 // CHECK5-NEXT: ret void
3711 //
3712 //
3713 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..12
3714 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3715 // CHECK5-NEXT: entry:
3716 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3717 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3718 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3719 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3720 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3721 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3722 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3723 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3724 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3725 // CHECK5-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3726 // CHECK5-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
3727 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3728 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3729 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3730 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3731 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3732 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3733 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3734 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3735 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3736 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3737 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3738 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3739 // CHECK5: cond.true:
3740 // CHECK5-NEXT: br label [[COND_END:%.*]]
3741 // CHECK5: cond.false:
3742 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3743 // CHECK5-NEXT: br label [[COND_END]]
3744 // CHECK5: cond.end:
3745 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3746 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3747 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3748 // CHECK5-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3749 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3750 // CHECK5: omp.inner.for.cond:
3751 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3752 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3753 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3754 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3755 // CHECK5: omp.inner.for.body:
3756 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3757 // CHECK5-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3758 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3759 // CHECK5-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3760 // CHECK5-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3761 // CHECK5-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3762 // CHECK5-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
3763 // CHECK5-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3764 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3765 // CHECK5: omp.inner.for.inc:
3766 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3767 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3768 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3769 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3770 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3771 // CHECK5: omp.inner.for.end:
3772 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3773 // CHECK5: omp.loop.exit:
3774 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3775 // CHECK5-NEXT: ret void
3776 //
3777 //
3778 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..13
3779 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3780 // CHECK5-NEXT: entry:
3781 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3782 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3783 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3784 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3785 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3786 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3787 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3788 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3789 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3790 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3791 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3792 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3793 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3794 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3795 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3796 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3797 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3798 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3799 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3800 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3801 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3802 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3803 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3804 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3805 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3806 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3807 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3808 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3809 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3810 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3811 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3812 // CHECK5: cond.true:
3813 // CHECK5-NEXT: br label [[COND_END:%.*]]
3814 // CHECK5: cond.false:
3815 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3816 // CHECK5-NEXT: br label [[COND_END]]
3817 // CHECK5: cond.end:
3818 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3819 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3820 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3821 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3822 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3823 // CHECK5: omp.inner.for.cond:
3824 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3825 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3826 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3827 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3828 // CHECK5: omp.inner.for.body:
3829 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3830 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3831 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3832 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3833 // CHECK5-NEXT: call void @_Z3fn2v()
3834 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3835 // CHECK5: omp.body.continue:
3836 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3837 // CHECK5: omp.inner.for.inc:
3838 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3839 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3840 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3841 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3842 // CHECK5: omp.inner.for.end:
3843 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3844 // CHECK5: omp.loop.exit:
3845 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3846 // CHECK5-NEXT: ret void
3847 //
3848 //
3849 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
3850 // CHECK5-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
3851 // CHECK5-NEXT: entry:
3852 // CHECK5-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
3853 // CHECK5-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
3854 // CHECK5-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
3855 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
3856 // CHECK5-NEXT: ret void
3857 //
3858 //
3859 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..14
3860 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
3861 // CHECK5-NEXT: entry:
3862 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3863 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3864 // CHECK5-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
3865 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3866 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3867 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3868 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3869 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3870 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3871 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3872 // CHECK5-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3873 // CHECK5-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
3874 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3875 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3876 // CHECK5-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
3877 // CHECK5-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
3878 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3879 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3880 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3881 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3882 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3883 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3884 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3885 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3886 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
3887 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3888 // CHECK5: cond.true:
3889 // CHECK5-NEXT: br label [[COND_END:%.*]]
3890 // CHECK5: cond.false:
3891 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3892 // CHECK5-NEXT: br label [[COND_END]]
3893 // CHECK5: cond.end:
3894 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3895 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3896 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3897 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3898 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3899 // CHECK5: omp.inner.for.cond:
3900 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3901 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3902 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3903 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3904 // CHECK5: omp.inner.for.body:
3905 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3906 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
3907 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3908 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3909 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
3910 // CHECK5-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
3911 // CHECK5-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3912 // CHECK5: omp_if.then:
3913 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
3914 // CHECK5-NEXT: br label [[OMP_IF_END:%.*]]
3915 // CHECK5: omp_if.else:
3916 // CHECK5-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
3917 // CHECK5-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3918 // CHECK5-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
3919 // CHECK5-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
3920 // CHECK5-NEXT: br label [[OMP_IF_END]]
3921 // CHECK5: omp_if.end:
3922 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3923 // CHECK5: omp.inner.for.inc:
3924 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3925 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3926 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
3927 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3928 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
3929 // CHECK5: omp.inner.for.end:
3930 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3931 // CHECK5: omp.loop.exit:
3932 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3933 // CHECK5-NEXT: ret void
3934 //
3935 //
3936 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..15
3937 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3938 // CHECK5-NEXT: entry:
3939 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3940 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3941 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3942 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3943 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3944 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3945 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3946 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3947 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3948 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3949 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3950 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3951 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3952 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3953 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3954 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3955 // CHECK5-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
3956 // CHECK5-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3957 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3958 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3959 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3960 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3961 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3962 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3963 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3964 // CHECK5-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3965 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3966 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3967 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3968 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3969 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3970 // CHECK5: cond.true:
3971 // CHECK5-NEXT: br label [[COND_END:%.*]]
3972 // CHECK5: cond.false:
3973 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3974 // CHECK5-NEXT: br label [[COND_END]]
3975 // CHECK5: cond.end:
3976 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3977 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3978 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3979 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3980 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3981 // CHECK5: omp.inner.for.cond:
3982 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3983 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3984 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3985 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3986 // CHECK5: omp.inner.for.body:
3987 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3988 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3989 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3990 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3991 // CHECK5-NEXT: call void @_Z3fn3v()
3992 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3993 // CHECK5: omp.body.continue:
3994 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3995 // CHECK5: omp.inner.for.inc:
3996 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3997 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3998 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3999 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4000 // CHECK5: omp.inner.for.end:
4001 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4002 // CHECK5: omp.loop.exit:
4003 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4004 // CHECK5-NEXT: ret void
4005 //
4006 //
4007 // CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4008 // CHECK5-SAME: () #[[ATTR5:[0-9]+]] {
4009 // CHECK5-NEXT: entry:
4010 // CHECK5-NEXT: call void @__tgt_register_requires(i64 1)
4011 // CHECK5-NEXT: ret void
4012 //
4013 //
4014 // CHECK6-LABEL: define {{[^@]+}}@_Z9gtid_testv
4015 // CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
4016 // CHECK6-NEXT: entry:
4017 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4018 // CHECK6-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
4019 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
4020 // CHECK6-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4021 // CHECK6-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4022 // CHECK6-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4023 // CHECK6: omp_offload.failed:
4024 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
4025 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT]]
4026 // CHECK6: omp_offload.cont:
4027 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4028 // CHECK6-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4029 // CHECK6-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4030 // CHECK6-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4031 // CHECK6: omp_offload.failed2:
4032 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
4033 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT3]]
4034 // CHECK6: omp_offload.cont3:
4035 // CHECK6-NEXT: ret void
4036 //
4037 //
4038 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
4039 // CHECK6-SAME: () #[[ATTR1:[0-9]+]] {
4040 // CHECK6-NEXT: entry:
4041 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
4042 // CHECK6-NEXT: ret void
4043 //
4044 //
4045 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined.
4046 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4047 // CHECK6-NEXT: entry:
4048 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4049 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4050 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4051 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4052 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4053 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4054 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4055 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4056 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4057 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4058 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4059 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4060 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4061 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4062 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4063 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4064 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4065 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4066 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4067 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4068 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4069 // CHECK6: cond.true:
4070 // CHECK6-NEXT: br label [[COND_END:%.*]]
4071 // CHECK6: cond.false:
4072 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4073 // CHECK6-NEXT: br label [[COND_END]]
4074 // CHECK6: cond.end:
4075 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4076 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4077 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4078 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4079 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4080 // CHECK6: omp.inner.for.cond:
4081 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4082 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4083 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4084 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4085 // CHECK6: omp.inner.for.body:
4086 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4087 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4088 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4089 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4090 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
4091 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4092 // CHECK6: omp.inner.for.inc:
4093 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4094 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4095 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4096 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4097 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4098 // CHECK6: omp.inner.for.end:
4099 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4100 // CHECK6: omp.loop.exit:
4101 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4102 // CHECK6-NEXT: ret void
4103 //
4104 //
4105 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..1
4106 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4107 // CHECK6-NEXT: entry:
4108 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4109 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4110 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4111 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4112 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4113 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4114 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4115 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4116 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4117 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4118 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4119 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4120 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4121 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4122 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4123 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4124 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4125 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4126 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4127 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4128 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4129 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4130 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4131 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4132 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4133 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4134 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4135 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4136 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4137 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4138 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4139 // CHECK6: cond.true:
4140 // CHECK6-NEXT: br label [[COND_END:%.*]]
4141 // CHECK6: cond.false:
4142 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4143 // CHECK6-NEXT: br label [[COND_END]]
4144 // CHECK6: cond.end:
4145 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4146 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4147 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4148 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4149 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4150 // CHECK6: omp.inner.for.cond:
4151 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4152 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4153 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4154 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4155 // CHECK6: omp.inner.for.body:
4156 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4157 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4158 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4159 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4160 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4161 // CHECK6: omp.body.continue:
4162 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4163 // CHECK6: omp.inner.for.inc:
4164 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4165 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4166 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4167 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4168 // CHECK6: omp.inner.for.end:
4169 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4170 // CHECK6: omp.loop.exit:
4171 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4172 // CHECK6-NEXT: ret void
4173 //
4174 //
4175 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
4176 // CHECK6-SAME: () #[[ATTR1]] {
4177 // CHECK6-NEXT: entry:
4178 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
4179 // CHECK6-NEXT: ret void
4180 //
4181 //
4182 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..2
4183 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4184 // CHECK6-NEXT: entry:
4185 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4186 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4187 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4188 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4189 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4190 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4191 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4192 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4193 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4194 // CHECK6-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4195 // CHECK6-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
4196 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4197 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4198 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4199 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4200 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4201 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4202 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4203 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4204 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4205 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4206 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4207 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4208 // CHECK6: cond.true:
4209 // CHECK6-NEXT: br label [[COND_END:%.*]]
4210 // CHECK6: cond.false:
4211 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4212 // CHECK6-NEXT: br label [[COND_END]]
4213 // CHECK6: cond.end:
4214 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4215 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4216 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4217 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4218 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4219 // CHECK6: omp.inner.for.cond:
4220 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4221 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4222 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4223 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4224 // CHECK6: omp.inner.for.body:
4225 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4226 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4227 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4228 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4229 // CHECK6-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4230 // CHECK6-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4231 // CHECK6-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
4232 // CHECK6-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4233 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4234 // CHECK6: omp.inner.for.inc:
4235 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4236 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4237 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4238 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4239 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4240 // CHECK6: omp.inner.for.end:
4241 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4242 // CHECK6: omp.loop.exit:
4243 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4244 // CHECK6-NEXT: ret void
4245 //
4246 //
4247 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..3
4248 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4249 // CHECK6-NEXT: entry:
4250 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4251 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4252 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4253 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4254 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4255 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4256 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4257 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4258 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4259 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4260 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4261 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4262 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4263 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4264 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4265 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4266 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4267 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4268 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4269 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4270 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4271 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4272 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4273 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4274 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4275 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4276 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4277 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4278 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4279 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4280 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4281 // CHECK6: cond.true:
4282 // CHECK6-NEXT: br label [[COND_END:%.*]]
4283 // CHECK6: cond.false:
4284 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4285 // CHECK6-NEXT: br label [[COND_END]]
4286 // CHECK6: cond.end:
4287 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4288 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4289 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4290 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4291 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4292 // CHECK6: omp.inner.for.cond:
4293 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4294 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4295 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4296 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4297 // CHECK6: omp.inner.for.body:
4298 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4299 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4300 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4301 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4302 // CHECK6-NEXT: call void @_Z9gtid_testv()
4303 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4304 // CHECK6: omp.body.continue:
4305 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4306 // CHECK6: omp.inner.for.inc:
4307 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4308 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4309 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4310 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4311 // CHECK6: omp.inner.for.end:
4312 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4313 // CHECK6: omp.loop.exit:
4314 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4315 // CHECK6-NEXT: ret void
4316 //
4317 //
4318 // CHECK6-LABEL: define {{[^@]+}}@main
4319 // CHECK6-SAME: () #[[ATTR3:[0-9]+]] {
4320 // CHECK6-NEXT: entry:
4321 // CHECK6-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
4322 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4323 // CHECK6-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
4324 // CHECK6-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
4325 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4326 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4327 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4328 // CHECK6-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
4329 // CHECK6-NEXT: store i32 0, i32* [[RETVAL]], align 4
4330 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4331 // CHECK6-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4332 // CHECK6-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4333 // CHECK6-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4334 // CHECK6: omp_offload.failed:
4335 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
4336 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT]]
4337 // CHECK6: omp_offload.cont:
4338 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4339 // CHECK6-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4340 // CHECK6-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4341 // CHECK6-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4342 // CHECK6: omp_offload.failed2:
4343 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
4344 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT3]]
4345 // CHECK6: omp_offload.cont3:
4346 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
4347 // CHECK6-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4348 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
4349 // CHECK6-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4350 // CHECK6-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4351 // CHECK6-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4352 // CHECK6-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
4353 // CHECK6-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4354 // CHECK6-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4355 // CHECK6-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
4356 // CHECK6-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4357 // CHECK6-NEXT: store i8* null, i8** [[TMP10]], align 8
4358 // CHECK6-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4359 // CHECK6-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4360 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4361 // CHECK6-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4362 // CHECK6-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4363 // CHECK6-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4364 // CHECK6: omp_offload.failed5:
4365 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
4366 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT6]]
4367 // CHECK6: omp_offload.cont6:
4368 // CHECK6-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
4369 // CHECK6-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
4370 // CHECK6-NEXT: ret i32 [[CALL]]
4371 //
4372 //
4373 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
4374 // CHECK6-SAME: () #[[ATTR1]] {
4375 // CHECK6-NEXT: entry:
4376 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
4377 // CHECK6-NEXT: ret void
4378 //
4379 //
4380 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..4
4381 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4382 // CHECK6-NEXT: entry:
4383 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4384 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4385 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4386 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4387 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4388 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4389 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4390 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4391 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4392 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4393 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4394 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4395 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4396 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4397 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4398 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4399 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4400 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4401 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4402 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4403 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4404 // CHECK6: cond.true:
4405 // CHECK6-NEXT: br label [[COND_END:%.*]]
4406 // CHECK6: cond.false:
4407 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4408 // CHECK6-NEXT: br label [[COND_END]]
4409 // CHECK6: cond.end:
4410 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4411 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4412 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4413 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4414 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4415 // CHECK6: omp.inner.for.cond:
4416 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4417 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4418 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4419 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4420 // CHECK6: omp.inner.for.body:
4421 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4422 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4423 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4424 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4425 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
4426 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4427 // CHECK6: omp.inner.for.inc:
4428 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4429 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4430 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4431 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4432 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4433 // CHECK6: omp.inner.for.end:
4434 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4435 // CHECK6: omp.loop.exit:
4436 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4437 // CHECK6-NEXT: ret void
4438 //
4439 //
4440 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..5
4441 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4442 // CHECK6-NEXT: entry:
4443 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4444 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4445 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4446 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4447 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4448 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4449 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4450 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4451 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4452 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4453 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4454 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4455 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4456 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4457 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4458 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4459 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4460 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4461 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4462 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4463 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4464 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4465 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4466 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4467 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4468 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4469 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4470 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4471 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4472 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4473 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4474 // CHECK6: cond.true:
4475 // CHECK6-NEXT: br label [[COND_END:%.*]]
4476 // CHECK6: cond.false:
4477 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4478 // CHECK6-NEXT: br label [[COND_END]]
4479 // CHECK6: cond.end:
4480 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4481 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4482 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4483 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4484 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4485 // CHECK6: omp.inner.for.cond:
4486 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4487 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4488 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4489 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4490 // CHECK6: omp.inner.for.body:
4491 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4492 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4493 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4494 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4495 // CHECK6-NEXT: call void @_Z3fn4v()
4496 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4497 // CHECK6: omp.body.continue:
4498 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4499 // CHECK6: omp.inner.for.inc:
4500 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4501 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4502 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4503 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4504 // CHECK6: omp.inner.for.end:
4505 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4506 // CHECK6: omp.loop.exit:
4507 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4508 // CHECK6-NEXT: ret void
4509 //
4510 //
4511 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
4512 // CHECK6-SAME: () #[[ATTR1]] {
4513 // CHECK6-NEXT: entry:
4514 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
4515 // CHECK6-NEXT: ret void
4516 //
4517 //
4518 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..6
4519 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4520 // CHECK6-NEXT: entry:
4521 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4522 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4523 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4524 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4525 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4526 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4527 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4528 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4529 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4530 // CHECK6-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4531 // CHECK6-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
4532 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4533 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4534 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4535 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4536 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4537 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4538 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4539 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4540 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4541 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4542 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4543 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4544 // CHECK6: cond.true:
4545 // CHECK6-NEXT: br label [[COND_END:%.*]]
4546 // CHECK6: cond.false:
4547 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4548 // CHECK6-NEXT: br label [[COND_END]]
4549 // CHECK6: cond.end:
4550 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4551 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4552 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4553 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4554 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4555 // CHECK6: omp.inner.for.cond:
4556 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4557 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4558 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4559 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4560 // CHECK6: omp.inner.for.body:
4561 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4562 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4563 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4564 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4565 // CHECK6-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4566 // CHECK6-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4567 // CHECK6-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
4568 // CHECK6-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4569 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4570 // CHECK6: omp.inner.for.inc:
4571 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4572 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4573 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4574 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4575 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4576 // CHECK6: omp.inner.for.end:
4577 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4578 // CHECK6: omp.loop.exit:
4579 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4580 // CHECK6-NEXT: ret void
4581 //
4582 //
4583 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..7
4584 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4585 // CHECK6-NEXT: entry:
4586 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4587 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4588 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4589 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4590 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4591 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4592 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4593 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4594 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4595 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4596 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4597 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4598 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4599 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4600 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4601 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4602 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4603 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4604 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4605 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4606 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4607 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4608 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4609 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4610 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4611 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4612 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4613 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4614 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4615 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4616 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4617 // CHECK6: cond.true:
4618 // CHECK6-NEXT: br label [[COND_END:%.*]]
4619 // CHECK6: cond.false:
4620 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4621 // CHECK6-NEXT: br label [[COND_END]]
4622 // CHECK6: cond.end:
4623 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4624 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4625 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4626 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4627 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4628 // CHECK6: omp.inner.for.cond:
4629 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4630 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4631 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4632 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4633 // CHECK6: omp.inner.for.body:
4634 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4635 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4636 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4637 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4638 // CHECK6-NEXT: call void @_Z3fn5v()
4639 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4640 // CHECK6: omp.body.continue:
4641 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4642 // CHECK6: omp.inner.for.inc:
4643 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4644 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4645 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4646 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4647 // CHECK6: omp.inner.for.end:
4648 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4649 // CHECK6: omp.loop.exit:
4650 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4651 // CHECK6-NEXT: ret void
4652 //
4653 //
4654 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
4655 // CHECK6-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
4656 // CHECK6-NEXT: entry:
4657 // CHECK6-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
4658 // CHECK6-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
4659 // CHECK6-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
4660 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
4661 // CHECK6-NEXT: ret void
4662 //
4663 //
4664 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..8
4665 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
4666 // CHECK6-NEXT: entry:
4667 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4668 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4669 // CHECK6-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
4670 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4671 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4672 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4673 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4674 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4675 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4676 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4677 // CHECK6-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4678 // CHECK6-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
4679 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4680 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4681 // CHECK6-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
4682 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
4683 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4684 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4685 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4686 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4687 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4688 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4689 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4690 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4691 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
4692 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4693 // CHECK6: cond.true:
4694 // CHECK6-NEXT: br label [[COND_END:%.*]]
4695 // CHECK6: cond.false:
4696 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4697 // CHECK6-NEXT: br label [[COND_END]]
4698 // CHECK6: cond.end:
4699 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4700 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4701 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4702 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4703 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4704 // CHECK6: omp.inner.for.cond:
4705 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4706 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4707 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4708 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4709 // CHECK6: omp.inner.for.body:
4710 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4711 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4712 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4713 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4714 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
4715 // CHECK6-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
4716 // CHECK6-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4717 // CHECK6: omp_if.then:
4718 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
4719 // CHECK6-NEXT: br label [[OMP_IF_END:%.*]]
4720 // CHECK6: omp_if.else:
4721 // CHECK6-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
4722 // CHECK6-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4723 // CHECK6-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
4724 // CHECK6-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
4725 // CHECK6-NEXT: br label [[OMP_IF_END]]
4726 // CHECK6: omp_if.end:
4727 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4728 // CHECK6: omp.inner.for.inc:
4729 // CHECK6-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4730 // CHECK6-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4731 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4732 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4733 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4734 // CHECK6: omp.inner.for.end:
4735 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4736 // CHECK6: omp.loop.exit:
4737 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4738 // CHECK6-NEXT: ret void
4739 //
4740 //
4741 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..9
4742 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4743 // CHECK6-NEXT: entry:
4744 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4745 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4746 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4747 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4748 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4749 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4750 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4751 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4752 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4753 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4754 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4755 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4756 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4757 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4758 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4759 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4760 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4761 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4762 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4763 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4764 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4765 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4766 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4767 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4768 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4769 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4770 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4771 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4772 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4773 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4774 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4775 // CHECK6: cond.true:
4776 // CHECK6-NEXT: br label [[COND_END:%.*]]
4777 // CHECK6: cond.false:
4778 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4779 // CHECK6-NEXT: br label [[COND_END]]
4780 // CHECK6: cond.end:
4781 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4782 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4783 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4784 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4785 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4786 // CHECK6: omp.inner.for.cond:
4787 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4788 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4789 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4790 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4791 // CHECK6: omp.inner.for.body:
4792 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4793 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4794 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4795 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4796 // CHECK6-NEXT: call void @_Z3fn6v()
4797 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4798 // CHECK6: omp.body.continue:
4799 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4800 // CHECK6: omp.inner.for.inc:
4801 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4802 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4803 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4804 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4805 // CHECK6: omp.inner.for.end:
4806 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4807 // CHECK6: omp.loop.exit:
4808 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4809 // CHECK6-NEXT: ret void
4810 //
4811 //
4812 // CHECK6-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
4813 // CHECK6-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
4814 // CHECK6-NEXT: entry:
4815 // CHECK6-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
4816 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4817 // CHECK6-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
4818 // CHECK6-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
4819 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4820 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4821 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4822 // CHECK6-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
4823 // CHECK6-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
4824 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4825 // CHECK6-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4826 // CHECK6-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4827 // CHECK6-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4828 // CHECK6: omp_offload.failed:
4829 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
4830 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT]]
4831 // CHECK6: omp_offload.cont:
4832 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4833 // CHECK6-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
4834 // CHECK6-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4835 // CHECK6-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4836 // CHECK6: omp_offload.failed2:
4837 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
4838 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT3]]
4839 // CHECK6: omp_offload.cont3:
4840 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
4841 // CHECK6-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4842 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
4843 // CHECK6-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4844 // CHECK6-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4845 // CHECK6-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4846 // CHECK6-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
4847 // CHECK6-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4848 // CHECK6-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4849 // CHECK6-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
4850 // CHECK6-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4851 // CHECK6-NEXT: store i8* null, i8** [[TMP10]], align 8
4852 // CHECK6-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4853 // CHECK6-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4854 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4855 // CHECK6-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4856 // CHECK6-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4857 // CHECK6-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4858 // CHECK6: omp_offload.failed5:
4859 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
4860 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT6]]
4861 // CHECK6: omp_offload.cont6:
4862 // CHECK6-NEXT: ret i32 0
4863 //
4864 //
4865 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
4866 // CHECK6-SAME: () #[[ATTR1]] {
4867 // CHECK6-NEXT: entry:
4868 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
4869 // CHECK6-NEXT: ret void
4870 //
4871 //
4872 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..10
4873 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4874 // CHECK6-NEXT: entry:
4875 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4876 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4877 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4878 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4879 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4880 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4881 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4882 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4883 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4884 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4885 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4886 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4887 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4888 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4889 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4890 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4891 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4892 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4893 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4894 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4895 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4896 // CHECK6: cond.true:
4897 // CHECK6-NEXT: br label [[COND_END:%.*]]
4898 // CHECK6: cond.false:
4899 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4900 // CHECK6-NEXT: br label [[COND_END]]
4901 // CHECK6: cond.end:
4902 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4903 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4904 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4905 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4906 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4907 // CHECK6: omp.inner.for.cond:
4908 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4909 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4910 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4911 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4912 // CHECK6: omp.inner.for.body:
4913 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4914 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4915 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4916 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4917 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
4918 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4919 // CHECK6: omp.inner.for.inc:
4920 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4921 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4922 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4923 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4924 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4925 // CHECK6: omp.inner.for.end:
4926 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4927 // CHECK6: omp.loop.exit:
4928 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4929 // CHECK6-NEXT: ret void
4930 //
4931 //
4932 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..11
4933 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4934 // CHECK6-NEXT: entry:
4935 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4936 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4937 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4938 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4939 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4940 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4941 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4942 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4943 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4944 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4945 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4946 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4947 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4948 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4949 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4950 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4951 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
4952 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4953 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4954 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4955 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4956 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4957 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4958 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4959 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4960 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4961 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4962 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4963 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4964 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4965 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4966 // CHECK6: cond.true:
4967 // CHECK6-NEXT: br label [[COND_END:%.*]]
4968 // CHECK6: cond.false:
4969 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4970 // CHECK6-NEXT: br label [[COND_END]]
4971 // CHECK6: cond.end:
4972 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4973 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4974 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4975 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4976 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4977 // CHECK6: omp.inner.for.cond:
4978 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4979 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4980 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4981 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4982 // CHECK6: omp.inner.for.body:
4983 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4984 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4985 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4986 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4987 // CHECK6-NEXT: call void @_Z3fn1v()
4988 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4989 // CHECK6: omp.body.continue:
4990 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4991 // CHECK6: omp.inner.for.inc:
4992 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4993 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4994 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4995 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4996 // CHECK6: omp.inner.for.end:
4997 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4998 // CHECK6: omp.loop.exit:
4999 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5000 // CHECK6-NEXT: ret void
5001 //
5002 //
5003 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
5004 // CHECK6-SAME: () #[[ATTR1]] {
5005 // CHECK6-NEXT: entry:
5006 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
5007 // CHECK6-NEXT: ret void
5008 //
5009 //
5010 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..12
5011 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5012 // CHECK6-NEXT: entry:
5013 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5014 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5015 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5016 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5017 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5018 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5019 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5020 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5021 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5022 // CHECK6-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5023 // CHECK6-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5024 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5025 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5026 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5027 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5028 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5029 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5030 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5031 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5032 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5033 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5034 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5035 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5036 // CHECK6: cond.true:
5037 // CHECK6-NEXT: br label [[COND_END:%.*]]
5038 // CHECK6: cond.false:
5039 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5040 // CHECK6-NEXT: br label [[COND_END]]
5041 // CHECK6: cond.end:
5042 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5043 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5044 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5045 // CHECK6-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5046 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5047 // CHECK6: omp.inner.for.cond:
5048 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5049 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5050 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5051 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5052 // CHECK6: omp.inner.for.body:
5053 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5054 // CHECK6-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5055 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5056 // CHECK6-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5057 // CHECK6-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5058 // CHECK6-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5059 // CHECK6-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5060 // CHECK6-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5061 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5062 // CHECK6: omp.inner.for.inc:
5063 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5064 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5065 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5066 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5067 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5068 // CHECK6: omp.inner.for.end:
5069 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5070 // CHECK6: omp.loop.exit:
5071 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5072 // CHECK6-NEXT: ret void
5073 //
5074 //
5075 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..13
5076 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5077 // CHECK6-NEXT: entry:
5078 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5079 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5080 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5081 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5082 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5083 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5084 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5085 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5086 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5087 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5088 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5089 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5090 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5091 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5092 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5093 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5094 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5095 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5096 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5097 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5098 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5099 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5100 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5101 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5102 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5103 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5104 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5105 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5106 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5107 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5108 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5109 // CHECK6: cond.true:
5110 // CHECK6-NEXT: br label [[COND_END:%.*]]
5111 // CHECK6: cond.false:
5112 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5113 // CHECK6-NEXT: br label [[COND_END]]
5114 // CHECK6: cond.end:
5115 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5116 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5117 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5118 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5119 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5120 // CHECK6: omp.inner.for.cond:
5121 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5122 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5123 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5124 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5125 // CHECK6: omp.inner.for.body:
5126 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5127 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5128 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5129 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5130 // CHECK6-NEXT: call void @_Z3fn2v()
5131 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5132 // CHECK6: omp.body.continue:
5133 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5134 // CHECK6: omp.inner.for.inc:
5135 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5136 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5137 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5138 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5139 // CHECK6: omp.inner.for.end:
5140 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5141 // CHECK6: omp.loop.exit:
5142 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5143 // CHECK6-NEXT: ret void
5144 //
5145 //
5146 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
5147 // CHECK6-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
5148 // CHECK6-NEXT: entry:
5149 // CHECK6-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
5150 // CHECK6-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
5151 // CHECK6-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
5152 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
5153 // CHECK6-NEXT: ret void
5154 //
5155 //
5156 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..14
5157 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
5158 // CHECK6-NEXT: entry:
5159 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5160 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5161 // CHECK6-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
5162 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5163 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5164 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5165 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5166 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5167 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5168 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5169 // CHECK6-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5170 // CHECK6-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5171 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5172 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5173 // CHECK6-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
5174 // CHECK6-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
5175 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5176 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5177 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5178 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5179 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5180 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5181 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5182 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5183 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
5184 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5185 // CHECK6: cond.true:
5186 // CHECK6-NEXT: br label [[COND_END:%.*]]
5187 // CHECK6: cond.false:
5188 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5189 // CHECK6-NEXT: br label [[COND_END]]
5190 // CHECK6: cond.end:
5191 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5192 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5193 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5194 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5195 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5196 // CHECK6: omp.inner.for.cond:
5197 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5198 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5199 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5200 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5201 // CHECK6: omp.inner.for.body:
5202 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5203 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5204 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5205 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5206 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
5207 // CHECK6-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
5208 // CHECK6-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5209 // CHECK6: omp_if.then:
5210 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
5211 // CHECK6-NEXT: br label [[OMP_IF_END:%.*]]
5212 // CHECK6: omp_if.else:
5213 // CHECK6-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
5214 // CHECK6-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5215 // CHECK6-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
5216 // CHECK6-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
5217 // CHECK6-NEXT: br label [[OMP_IF_END]]
5218 // CHECK6: omp_if.end:
5219 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5220 // CHECK6: omp.inner.for.inc:
5221 // CHECK6-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5222 // CHECK6-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5223 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
5224 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5225 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5226 // CHECK6: omp.inner.for.end:
5227 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5228 // CHECK6: omp.loop.exit:
5229 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5230 // CHECK6-NEXT: ret void
5231 //
5232 //
5233 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..15
5234 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5235 // CHECK6-NEXT: entry:
5236 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5237 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5238 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5239 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5240 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5241 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5242 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5243 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5244 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5245 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5246 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5247 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5248 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5249 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5250 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5251 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5252 // CHECK6-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5253 // CHECK6-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5254 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5255 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5256 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5257 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5258 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5259 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5260 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5261 // CHECK6-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5262 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5263 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5264 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5265 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5266 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5267 // CHECK6: cond.true:
5268 // CHECK6-NEXT: br label [[COND_END:%.*]]
5269 // CHECK6: cond.false:
5270 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5271 // CHECK6-NEXT: br label [[COND_END]]
5272 // CHECK6: cond.end:
5273 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5274 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5275 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5276 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5277 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5278 // CHECK6: omp.inner.for.cond:
5279 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5280 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5281 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5282 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5283 // CHECK6: omp.inner.for.body:
5284 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5285 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5286 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5287 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5288 // CHECK6-NEXT: call void @_Z3fn3v()
5289 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5290 // CHECK6: omp.body.continue:
5291 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5292 // CHECK6: omp.inner.for.inc:
5293 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5294 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5295 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5296 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5297 // CHECK6: omp.inner.for.end:
5298 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5299 // CHECK6: omp.loop.exit:
5300 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5301 // CHECK6-NEXT: ret void
5302 //
5303 //
5304 // CHECK6-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5305 // CHECK6-SAME: () #[[ATTR5:[0-9]+]] {
5306 // CHECK6-NEXT: entry:
5307 // CHECK6-NEXT: call void @__tgt_register_requires(i64 1)
5308 // CHECK6-NEXT: ret void
5309 //
5310 //
5311 // CHECK9-LABEL: define {{[^@]+}}@_Z9gtid_testv
5312 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
5313 // CHECK9-NEXT: entry:
5314 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5315 // CHECK9-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
5316 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
5317 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
5318 // CHECK9-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
5319 // CHECK9-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5320 // CHECK9: omp_offload.failed:
5321 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
5322 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
5323 // CHECK9: omp_offload.cont:
5324 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5325 // CHECK9-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
5326 // CHECK9-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
5327 // CHECK9-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
5328 // CHECK9: omp_offload.failed2:
5329 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
5330 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT3]]
5331 // CHECK9: omp_offload.cont3:
5332 // CHECK9-NEXT: ret void
5333 //
5334 //
5335 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
5336 // CHECK9-SAME: () #[[ATTR1:[0-9]+]] {
5337 // CHECK9-NEXT: entry:
5338 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
5339 // CHECK9-NEXT: ret void
5340 //
5341 //
5342 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
5343 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5344 // CHECK9-NEXT: entry:
5345 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5346 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5347 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5348 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5349 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5350 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5351 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5352 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5353 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5354 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5355 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5356 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5357 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5358 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5359 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5360 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5361 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5362 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5363 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5364 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5365 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5366 // CHECK9: cond.true:
5367 // CHECK9-NEXT: br label [[COND_END:%.*]]
5368 // CHECK9: cond.false:
5369 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5370 // CHECK9-NEXT: br label [[COND_END]]
5371 // CHECK9: cond.end:
5372 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5373 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5374 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5375 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5376 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5377 // CHECK9: omp.inner.for.cond:
5378 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5379 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5380 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5381 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5382 // CHECK9: omp.inner.for.body:
5383 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5384 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5385 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5386 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5387 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
5388 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5389 // CHECK9: omp.inner.for.inc:
5390 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5391 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5392 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5393 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5394 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5395 // CHECK9: omp.inner.for.end:
5396 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5397 // CHECK9: omp.loop.exit:
5398 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5399 // CHECK9-NEXT: ret void
5400 //
5401 //
5402 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
5403 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5404 // CHECK9-NEXT: entry:
5405 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5406 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5407 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5408 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5409 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5410 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5411 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5412 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5413 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5414 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5415 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5416 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5417 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5418 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5419 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5420 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5421 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5422 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5423 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5424 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5425 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5426 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5427 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5428 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5429 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5430 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5431 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5432 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5433 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5434 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5435 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5436 // CHECK9: cond.true:
5437 // CHECK9-NEXT: br label [[COND_END:%.*]]
5438 // CHECK9: cond.false:
5439 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5440 // CHECK9-NEXT: br label [[COND_END]]
5441 // CHECK9: cond.end:
5442 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5443 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5444 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5445 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5446 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5447 // CHECK9: omp.inner.for.cond:
5448 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5449 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5450 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5451 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5452 // CHECK9: omp.inner.for.body:
5453 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5454 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5455 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5456 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5457 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5458 // CHECK9: omp.body.continue:
5459 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5460 // CHECK9: omp.inner.for.inc:
5461 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5462 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5463 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5464 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5465 // CHECK9: omp.inner.for.end:
5466 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5467 // CHECK9: omp.loop.exit:
5468 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5469 // CHECK9-NEXT: ret void
5470 //
5471 //
5472 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
5473 // CHECK9-SAME: () #[[ATTR1]] {
5474 // CHECK9-NEXT: entry:
5475 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
5476 // CHECK9-NEXT: ret void
5477 //
5478 //
5479 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
5480 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5481 // CHECK9-NEXT: entry:
5482 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5483 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5484 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5485 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5486 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5487 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5488 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5489 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5490 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5491 // CHECK9-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5492 // CHECK9-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5493 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5494 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5495 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5496 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5497 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5498 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5499 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5500 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5501 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5502 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5503 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5504 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5505 // CHECK9: cond.true:
5506 // CHECK9-NEXT: br label [[COND_END:%.*]]
5507 // CHECK9: cond.false:
5508 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5509 // CHECK9-NEXT: br label [[COND_END]]
5510 // CHECK9: cond.end:
5511 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5512 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5513 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5514 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5515 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5516 // CHECK9: omp.inner.for.cond:
5517 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5518 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5519 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5520 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5521 // CHECK9: omp.inner.for.body:
5522 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5523 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5524 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5525 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5526 // CHECK9-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5527 // CHECK9-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5528 // CHECK9-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5529 // CHECK9-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5530 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5531 // CHECK9: omp.inner.for.inc:
5532 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5533 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5534 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5535 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5536 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5537 // CHECK9: omp.inner.for.end:
5538 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5539 // CHECK9: omp.loop.exit:
5540 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5541 // CHECK9-NEXT: ret void
5542 //
5543 //
5544 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
5545 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5546 // CHECK9-NEXT: entry:
5547 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5548 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5549 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5550 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5551 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5552 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5553 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5554 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5555 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5556 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5557 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5558 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5559 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5560 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5561 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5562 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5563 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5564 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5565 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5566 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5567 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5568 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5569 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5570 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5571 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5572 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5573 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5574 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5575 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5576 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5577 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5578 // CHECK9: cond.true:
5579 // CHECK9-NEXT: br label [[COND_END:%.*]]
5580 // CHECK9: cond.false:
5581 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5582 // CHECK9-NEXT: br label [[COND_END]]
5583 // CHECK9: cond.end:
5584 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5585 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5586 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5587 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5588 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5589 // CHECK9: omp.inner.for.cond:
5590 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5591 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5592 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5593 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5594 // CHECK9: omp.inner.for.body:
5595 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5596 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5597 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5598 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5599 // CHECK9-NEXT: call void @_Z9gtid_testv()
5600 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5601 // CHECK9: omp.body.continue:
5602 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5603 // CHECK9: omp.inner.for.inc:
5604 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5605 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5606 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5607 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5608 // CHECK9: omp.inner.for.end:
5609 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5610 // CHECK9: omp.loop.exit:
5611 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5612 // CHECK9-NEXT: ret void
5613 //
5614 //
5615 // CHECK9-LABEL: define {{[^@]+}}@main
5616 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
5617 // CHECK9-NEXT: entry:
5618 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
5619 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5620 // CHECK9-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
5621 // CHECK9-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
5622 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
5623 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
5624 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
5625 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
5626 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
5627 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5628 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
5629 // CHECK9-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
5630 // CHECK9-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5631 // CHECK9: omp_offload.failed:
5632 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
5633 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
5634 // CHECK9: omp_offload.cont:
5635 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5636 // CHECK9-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
5637 // CHECK9-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
5638 // CHECK9-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
5639 // CHECK9: omp_offload.failed2:
5640 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
5641 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT3]]
5642 // CHECK9: omp_offload.cont3:
5643 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
5644 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
5645 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
5646 // CHECK9-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
5647 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5648 // CHECK9-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
5649 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
5650 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5651 // CHECK9-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
5652 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
5653 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
5654 // CHECK9-NEXT: store i8* null, i8** [[TMP10]], align 8
5655 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5656 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5657 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5658 // CHECK9-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5659 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5660 // CHECK9-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
5661 // CHECK9: omp_offload.failed5:
5662 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
5663 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT6]]
5664 // CHECK9: omp_offload.cont6:
5665 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
5666 // CHECK9-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
5667 // CHECK9-NEXT: ret i32 [[CALL]]
5668 //
5669 //
5670 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
5671 // CHECK9-SAME: () #[[ATTR1]] {
5672 // CHECK9-NEXT: entry:
5673 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
5674 // CHECK9-NEXT: ret void
5675 //
5676 //
5677 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
5678 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5679 // CHECK9-NEXT: entry:
5680 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5681 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5682 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5683 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5684 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5685 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5686 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5687 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5688 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5689 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5690 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5691 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5692 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5693 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5694 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5695 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5696 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5697 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5698 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5699 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5700 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5701 // CHECK9: cond.true:
5702 // CHECK9-NEXT: br label [[COND_END:%.*]]
5703 // CHECK9: cond.false:
5704 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5705 // CHECK9-NEXT: br label [[COND_END]]
5706 // CHECK9: cond.end:
5707 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5708 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5709 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5710 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5711 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5712 // CHECK9: omp.inner.for.cond:
5713 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5714 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5715 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5716 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5717 // CHECK9: omp.inner.for.body:
5718 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5719 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5720 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5721 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5722 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
5723 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5724 // CHECK9: omp.inner.for.inc:
5725 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5726 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5727 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5728 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5729 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5730 // CHECK9: omp.inner.for.end:
5731 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5732 // CHECK9: omp.loop.exit:
5733 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5734 // CHECK9-NEXT: ret void
5735 //
5736 //
5737 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
5738 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5739 // CHECK9-NEXT: entry:
5740 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5741 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5742 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5743 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5744 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5745 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5746 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5747 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5748 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5749 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5750 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5751 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5752 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5753 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5754 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5755 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5756 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5757 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5758 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5759 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5760 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5761 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5762 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5763 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5764 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5765 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5766 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5767 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5768 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5769 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5770 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5771 // CHECK9: cond.true:
5772 // CHECK9-NEXT: br label [[COND_END:%.*]]
5773 // CHECK9: cond.false:
5774 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5775 // CHECK9-NEXT: br label [[COND_END]]
5776 // CHECK9: cond.end:
5777 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5778 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5779 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5780 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5781 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5782 // CHECK9: omp.inner.for.cond:
5783 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5784 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5785 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5786 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5787 // CHECK9: omp.inner.for.body:
5788 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5789 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5790 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5791 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5792 // CHECK9-NEXT: call void @_Z3fn4v()
5793 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5794 // CHECK9: omp.body.continue:
5795 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5796 // CHECK9: omp.inner.for.inc:
5797 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5798 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5799 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5800 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5801 // CHECK9: omp.inner.for.end:
5802 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5803 // CHECK9: omp.loop.exit:
5804 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5805 // CHECK9-NEXT: ret void
5806 //
5807 //
5808 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
5809 // CHECK9-SAME: () #[[ATTR1]] {
5810 // CHECK9-NEXT: entry:
5811 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
5812 // CHECK9-NEXT: ret void
5813 //
5814 //
5815 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
5816 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5817 // CHECK9-NEXT: entry:
5818 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5819 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5820 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5821 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5822 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5823 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5824 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5825 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5826 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5827 // CHECK9-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5828 // CHECK9-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5829 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5830 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5831 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5832 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5833 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5834 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5835 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5836 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5837 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5838 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5839 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5840 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5841 // CHECK9: cond.true:
5842 // CHECK9-NEXT: br label [[COND_END:%.*]]
5843 // CHECK9: cond.false:
5844 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5845 // CHECK9-NEXT: br label [[COND_END]]
5846 // CHECK9: cond.end:
5847 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5848 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5849 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5850 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5851 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5852 // CHECK9: omp.inner.for.cond:
5853 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5854 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5855 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5856 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5857 // CHECK9: omp.inner.for.body:
5858 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5859 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5860 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5861 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5862 // CHECK9-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5863 // CHECK9-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5864 // CHECK9-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5865 // CHECK9-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5866 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5867 // CHECK9: omp.inner.for.inc:
5868 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5869 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5870 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5871 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5872 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5873 // CHECK9: omp.inner.for.end:
5874 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5875 // CHECK9: omp.loop.exit:
5876 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5877 // CHECK9-NEXT: ret void
5878 //
5879 //
5880 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
5881 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5882 // CHECK9-NEXT: entry:
5883 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5884 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5885 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5886 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5887 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5888 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5889 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5890 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5891 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5892 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5893 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5894 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5895 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5896 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5897 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5898 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5899 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
5900 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5901 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5902 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5903 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5904 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5905 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5906 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5907 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5908 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5909 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5910 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5911 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5912 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5913 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5914 // CHECK9: cond.true:
5915 // CHECK9-NEXT: br label [[COND_END:%.*]]
5916 // CHECK9: cond.false:
5917 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5918 // CHECK9-NEXT: br label [[COND_END]]
5919 // CHECK9: cond.end:
5920 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5921 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5922 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5923 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5924 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5925 // CHECK9: omp.inner.for.cond:
5926 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5927 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5928 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5929 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5930 // CHECK9: omp.inner.for.body:
5931 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5932 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5933 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5934 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5935 // CHECK9-NEXT: call void @_Z3fn5v()
5936 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5937 // CHECK9: omp.body.continue:
5938 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5939 // CHECK9: omp.inner.for.inc:
5940 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5941 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5942 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5943 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
5944 // CHECK9: omp.inner.for.end:
5945 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5946 // CHECK9: omp.loop.exit:
5947 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5948 // CHECK9-NEXT: ret void
5949 //
5950 //
5951 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
5952 // CHECK9-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
5953 // CHECK9-NEXT: entry:
5954 // CHECK9-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
5955 // CHECK9-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
5956 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
5957 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
5958 // CHECK9-NEXT: ret void
5959 //
5960 //
5961 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..8
5962 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
5963 // CHECK9-NEXT: entry:
5964 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5965 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5966 // CHECK9-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
5967 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5968 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
5969 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5970 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5971 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5972 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5973 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
5974 // CHECK9-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5975 // CHECK9-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5976 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5977 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5978 // CHECK9-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
5979 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
5980 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5981 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5982 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5983 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5984 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5985 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5986 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5987 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5988 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
5989 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5990 // CHECK9: cond.true:
5991 // CHECK9-NEXT: br label [[COND_END:%.*]]
5992 // CHECK9: cond.false:
5993 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5994 // CHECK9-NEXT: br label [[COND_END]]
5995 // CHECK9: cond.end:
5996 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5997 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5998 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5999 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6000 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6001 // CHECK9: omp.inner.for.cond:
6002 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6003 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6004 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6005 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6006 // CHECK9: omp.inner.for.body:
6007 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6008 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
6009 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6010 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
6011 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
6012 // CHECK9-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
6013 // CHECK9-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6014 // CHECK9: omp_if.then:
6015 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
6016 // CHECK9-NEXT: br label [[OMP_IF_END:%.*]]
6017 // CHECK9: omp_if.else:
6018 // CHECK9-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
6019 // CHECK9-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6020 // CHECK9-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
6021 // CHECK9-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
6022 // CHECK9-NEXT: br label [[OMP_IF_END]]
6023 // CHECK9: omp_if.end:
6024 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6025 // CHECK9: omp.inner.for.inc:
6026 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6027 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6028 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6029 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6030 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6031 // CHECK9: omp.inner.for.end:
6032 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6033 // CHECK9: omp.loop.exit:
6034 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6035 // CHECK9-NEXT: ret void
6036 //
6037 //
6038 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..9
6039 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6040 // CHECK9-NEXT: entry:
6041 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6042 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6043 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6044 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6045 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6046 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6047 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6048 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6049 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6050 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6051 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6052 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6053 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6054 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6055 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6056 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6057 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6058 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6059 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6060 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6061 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6062 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6063 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6064 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6065 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6066 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6067 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6068 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6069 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6070 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6071 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6072 // CHECK9: cond.true:
6073 // CHECK9-NEXT: br label [[COND_END:%.*]]
6074 // CHECK9: cond.false:
6075 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6076 // CHECK9-NEXT: br label [[COND_END]]
6077 // CHECK9: cond.end:
6078 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6079 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6080 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6081 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6082 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6083 // CHECK9: omp.inner.for.cond:
6084 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6085 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6086 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6087 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6088 // CHECK9: omp.inner.for.body:
6089 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6090 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6091 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6092 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6093 // CHECK9-NEXT: call void @_Z3fn6v()
6094 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6095 // CHECK9: omp.body.continue:
6096 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6097 // CHECK9: omp.inner.for.inc:
6098 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6099 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6100 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6101 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6102 // CHECK9: omp.inner.for.end:
6103 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6104 // CHECK9: omp.loop.exit:
6105 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6106 // CHECK9-NEXT: ret void
6107 //
6108 //
6109 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6110 // CHECK9-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
6111 // CHECK9-NEXT: entry:
6112 // CHECK9-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
6113 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6114 // CHECK9-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
6115 // CHECK9-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
6116 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
6117 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
6118 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
6119 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
6120 // CHECK9-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
6121 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6122 // CHECK9-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6123 // CHECK9-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
6124 // CHECK9-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6125 // CHECK9: omp_offload.failed:
6126 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
6127 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
6128 // CHECK9: omp_offload.cont:
6129 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6130 // CHECK9-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6131 // CHECK9-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
6132 // CHECK9-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
6133 // CHECK9: omp_offload.failed2:
6134 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
6135 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT3]]
6136 // CHECK9: omp_offload.cont3:
6137 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
6138 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
6139 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
6140 // CHECK9-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
6141 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6142 // CHECK9-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
6143 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
6144 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6145 // CHECK9-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
6146 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
6147 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
6148 // CHECK9-NEXT: store i8* null, i8** [[TMP10]], align 8
6149 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6150 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6151 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6152 // CHECK9-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6153 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6154 // CHECK9-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
6155 // CHECK9: omp_offload.failed5:
6156 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
6157 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT6]]
6158 // CHECK9: omp_offload.cont6:
6159 // CHECK9-NEXT: ret i32 0
6160 //
6161 //
6162 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
6163 // CHECK9-SAME: () #[[ATTR1]] {
6164 // CHECK9-NEXT: entry:
6165 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
6166 // CHECK9-NEXT: ret void
6167 //
6168 //
6169 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
6170 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6171 // CHECK9-NEXT: entry:
6172 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6173 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6174 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6175 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6176 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6177 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6178 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6179 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6180 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6181 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6182 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6183 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6184 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6185 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6186 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6187 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6188 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6189 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6190 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6191 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6192 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6193 // CHECK9: cond.true:
6194 // CHECK9-NEXT: br label [[COND_END:%.*]]
6195 // CHECK9: cond.false:
6196 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6197 // CHECK9-NEXT: br label [[COND_END]]
6198 // CHECK9: cond.end:
6199 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6200 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6201 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6202 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6203 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6204 // CHECK9: omp.inner.for.cond:
6205 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6206 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6207 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6208 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6209 // CHECK9: omp.inner.for.body:
6210 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6211 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6212 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6213 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6214 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
6215 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6216 // CHECK9: omp.inner.for.inc:
6217 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6218 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6219 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
6220 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6221 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6222 // CHECK9: omp.inner.for.end:
6223 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6224 // CHECK9: omp.loop.exit:
6225 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6226 // CHECK9-NEXT: ret void
6227 //
6228 //
6229 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
6230 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6231 // CHECK9-NEXT: entry:
6232 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6233 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6234 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6235 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6236 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6237 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6238 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6239 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6240 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6241 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6242 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6243 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6244 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6245 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6246 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6247 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6248 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6249 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6250 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6251 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6252 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6253 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6254 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6255 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6256 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6257 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6258 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6259 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6260 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6261 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6262 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6263 // CHECK9: cond.true:
6264 // CHECK9-NEXT: br label [[COND_END:%.*]]
6265 // CHECK9: cond.false:
6266 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6267 // CHECK9-NEXT: br label [[COND_END]]
6268 // CHECK9: cond.end:
6269 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6270 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6271 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6272 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6273 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6274 // CHECK9: omp.inner.for.cond:
6275 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6276 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6277 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6278 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6279 // CHECK9: omp.inner.for.body:
6280 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6281 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6282 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6283 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6284 // CHECK9-NEXT: call void @_Z3fn1v()
6285 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6286 // CHECK9: omp.body.continue:
6287 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6288 // CHECK9: omp.inner.for.inc:
6289 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6290 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6291 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6292 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6293 // CHECK9: omp.inner.for.end:
6294 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6295 // CHECK9: omp.loop.exit:
6296 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6297 // CHECK9-NEXT: ret void
6298 //
6299 //
6300 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
6301 // CHECK9-SAME: () #[[ATTR1]] {
6302 // CHECK9-NEXT: entry:
6303 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
6304 // CHECK9-NEXT: ret void
6305 //
6306 //
6307 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..12
6308 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6309 // CHECK9-NEXT: entry:
6310 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6311 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6312 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6313 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6314 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6315 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6316 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6317 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6318 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6319 // CHECK9-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6320 // CHECK9-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
6321 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6322 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6323 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6324 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6325 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6326 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6327 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6328 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6329 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6330 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6331 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6332 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6333 // CHECK9: cond.true:
6334 // CHECK9-NEXT: br label [[COND_END:%.*]]
6335 // CHECK9: cond.false:
6336 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6337 // CHECK9-NEXT: br label [[COND_END]]
6338 // CHECK9: cond.end:
6339 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6340 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6341 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6342 // CHECK9-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6343 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6344 // CHECK9: omp.inner.for.cond:
6345 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6346 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6347 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6348 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6349 // CHECK9: omp.inner.for.body:
6350 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6351 // CHECK9-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6352 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6353 // CHECK9-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6354 // CHECK9-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6355 // CHECK9-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6356 // CHECK9-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
6357 // CHECK9-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6358 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6359 // CHECK9: omp.inner.for.inc:
6360 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6361 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6362 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
6363 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6364 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6365 // CHECK9: omp.inner.for.end:
6366 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6367 // CHECK9: omp.loop.exit:
6368 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6369 // CHECK9-NEXT: ret void
6370 //
6371 //
6372 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..13
6373 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6374 // CHECK9-NEXT: entry:
6375 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6376 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6377 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6378 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6379 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6380 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6381 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6382 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6383 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6384 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6385 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6386 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6387 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6388 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6389 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6390 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6391 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6392 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6393 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6394 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6395 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6396 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6397 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6398 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6399 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6400 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6401 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6402 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6403 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6404 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6405 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6406 // CHECK9: cond.true:
6407 // CHECK9-NEXT: br label [[COND_END:%.*]]
6408 // CHECK9: cond.false:
6409 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6410 // CHECK9-NEXT: br label [[COND_END]]
6411 // CHECK9: cond.end:
6412 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6413 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6414 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6415 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6416 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6417 // CHECK9: omp.inner.for.cond:
6418 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6419 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6420 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6421 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6422 // CHECK9: omp.inner.for.body:
6423 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6424 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6425 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6426 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6427 // CHECK9-NEXT: call void @_Z3fn2v()
6428 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6429 // CHECK9: omp.body.continue:
6430 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6431 // CHECK9: omp.inner.for.inc:
6432 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6433 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6434 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6435 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6436 // CHECK9: omp.inner.for.end:
6437 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6438 // CHECK9: omp.loop.exit:
6439 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6440 // CHECK9-NEXT: ret void
6441 //
6442 //
6443 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
6444 // CHECK9-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
6445 // CHECK9-NEXT: entry:
6446 // CHECK9-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
6447 // CHECK9-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
6448 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
6449 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
6450 // CHECK9-NEXT: ret void
6451 //
6452 //
6453 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..14
6454 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
6455 // CHECK9-NEXT: entry:
6456 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6457 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6458 // CHECK9-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
6459 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6460 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6461 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6462 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6463 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6464 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6465 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6466 // CHECK9-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6467 // CHECK9-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
6468 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6469 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6470 // CHECK9-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
6471 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
6472 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6473 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6474 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6475 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6476 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6477 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6478 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6479 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6480 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
6481 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6482 // CHECK9: cond.true:
6483 // CHECK9-NEXT: br label [[COND_END:%.*]]
6484 // CHECK9: cond.false:
6485 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6486 // CHECK9-NEXT: br label [[COND_END]]
6487 // CHECK9: cond.end:
6488 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6489 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6490 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6491 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6492 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6493 // CHECK9: omp.inner.for.cond:
6494 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6495 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6496 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6497 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6498 // CHECK9: omp.inner.for.body:
6499 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6500 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
6501 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6502 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
6503 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
6504 // CHECK9-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
6505 // CHECK9-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6506 // CHECK9: omp_if.then:
6507 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
6508 // CHECK9-NEXT: br label [[OMP_IF_END:%.*]]
6509 // CHECK9: omp_if.else:
6510 // CHECK9-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
6511 // CHECK9-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6512 // CHECK9-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
6513 // CHECK9-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
6514 // CHECK9-NEXT: br label [[OMP_IF_END]]
6515 // CHECK9: omp_if.end:
6516 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6517 // CHECK9: omp.inner.for.inc:
6518 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6519 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6520 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6521 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6522 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6523 // CHECK9: omp.inner.for.end:
6524 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6525 // CHECK9: omp.loop.exit:
6526 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6527 // CHECK9-NEXT: ret void
6528 //
6529 //
6530 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
6531 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6532 // CHECK9-NEXT: entry:
6533 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6534 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6535 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6536 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6537 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6538 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
6539 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6540 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6541 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6542 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6543 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
6544 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6545 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6546 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6547 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6548 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6549 // CHECK9-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6550 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6551 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6552 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6553 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6554 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6555 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6556 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6557 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6558 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6559 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6560 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6561 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6562 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6563 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6564 // CHECK9: cond.true:
6565 // CHECK9-NEXT: br label [[COND_END:%.*]]
6566 // CHECK9: cond.false:
6567 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6568 // CHECK9-NEXT: br label [[COND_END]]
6569 // CHECK9: cond.end:
6570 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6571 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6572 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6573 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6574 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6575 // CHECK9: omp.inner.for.cond:
6576 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6577 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6578 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6579 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6580 // CHECK9: omp.inner.for.body:
6581 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6582 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6583 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6584 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6585 // CHECK9-NEXT: call void @_Z3fn3v()
6586 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6587 // CHECK9: omp.body.continue:
6588 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6589 // CHECK9: omp.inner.for.inc:
6590 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6591 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6592 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6593 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
6594 // CHECK9: omp.inner.for.end:
6595 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6596 // CHECK9: omp.loop.exit:
6597 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6598 // CHECK9-NEXT: ret void
6599 //
6600 //
6601 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6602 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
6603 // CHECK9-NEXT: entry:
6604 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
6605 // CHECK9-NEXT: ret void
6606 //
6607 //
6608 // CHECK10-LABEL: define {{[^@]+}}@_Z9gtid_testv
6609 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
6610 // CHECK10-NEXT: entry:
6611 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6612 // CHECK10-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
6613 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
6614 // CHECK10-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6615 // CHECK10-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
6616 // CHECK10-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6617 // CHECK10: omp_offload.failed:
6618 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
6619 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
6620 // CHECK10: omp_offload.cont:
6621 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6622 // CHECK10-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6623 // CHECK10-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
6624 // CHECK10-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
6625 // CHECK10: omp_offload.failed2:
6626 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
6627 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT3]]
6628 // CHECK10: omp_offload.cont3:
6629 // CHECK10-NEXT: ret void
6630 //
6631 //
6632 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
6633 // CHECK10-SAME: () #[[ATTR1:[0-9]+]] {
6634 // CHECK10-NEXT: entry:
6635 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
6636 // CHECK10-NEXT: ret void
6637 //
6638 //
6639 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
6640 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6641 // CHECK10-NEXT: entry:
6642 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6643 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6644 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6645 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6646 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6647 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6648 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6649 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6650 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6651 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6652 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6653 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6654 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6655 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6656 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6657 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6658 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6659 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6660 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6661 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6662 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6663 // CHECK10: cond.true:
6664 // CHECK10-NEXT: br label [[COND_END:%.*]]
6665 // CHECK10: cond.false:
6666 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6667 // CHECK10-NEXT: br label [[COND_END]]
6668 // CHECK10: cond.end:
6669 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6670 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6671 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6672 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6673 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6674 // CHECK10: omp.inner.for.cond:
6675 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6676 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6677 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6678 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6679 // CHECK10: omp.inner.for.body:
6680 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6681 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6682 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6683 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6684 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
6685 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6686 // CHECK10: omp.inner.for.inc:
6687 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6688 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6689 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
6690 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6691 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
6692 // CHECK10: omp.inner.for.end:
6693 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6694 // CHECK10: omp.loop.exit:
6695 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6696 // CHECK10-NEXT: ret void
6697 //
6698 //
6699 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
6700 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6701 // CHECK10-NEXT: entry:
6702 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6703 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6704 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6705 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6706 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6707 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6708 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6709 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6710 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6711 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6712 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6713 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6714 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6715 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6716 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6717 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6718 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6719 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6720 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6721 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6722 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6723 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6724 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6725 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6726 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6727 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6728 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6729 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6730 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6731 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6732 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6733 // CHECK10: cond.true:
6734 // CHECK10-NEXT: br label [[COND_END:%.*]]
6735 // CHECK10: cond.false:
6736 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6737 // CHECK10-NEXT: br label [[COND_END]]
6738 // CHECK10: cond.end:
6739 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6740 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6741 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6742 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6743 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6744 // CHECK10: omp.inner.for.cond:
6745 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6746 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6747 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6748 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6749 // CHECK10: omp.inner.for.body:
6750 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6751 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6752 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6753 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6754 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6755 // CHECK10: omp.body.continue:
6756 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6757 // CHECK10: omp.inner.for.inc:
6758 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6759 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6760 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6761 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
6762 // CHECK10: omp.inner.for.end:
6763 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6764 // CHECK10: omp.loop.exit:
6765 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6766 // CHECK10-NEXT: ret void
6767 //
6768 //
6769 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
6770 // CHECK10-SAME: () #[[ATTR1]] {
6771 // CHECK10-NEXT: entry:
6772 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
6773 // CHECK10-NEXT: ret void
6774 //
6775 //
6776 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
6777 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6778 // CHECK10-NEXT: entry:
6779 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6780 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6781 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6782 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6783 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6784 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6785 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6786 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6787 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6788 // CHECK10-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6789 // CHECK10-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
6790 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6791 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6792 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6793 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6794 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6795 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6796 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6797 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6798 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6799 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6800 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6801 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6802 // CHECK10: cond.true:
6803 // CHECK10-NEXT: br label [[COND_END:%.*]]
6804 // CHECK10: cond.false:
6805 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6806 // CHECK10-NEXT: br label [[COND_END]]
6807 // CHECK10: cond.end:
6808 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6809 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6810 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6811 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6812 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6813 // CHECK10: omp.inner.for.cond:
6814 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6815 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6816 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6817 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6818 // CHECK10: omp.inner.for.body:
6819 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6820 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6821 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6822 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6823 // CHECK10-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6824 // CHECK10-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6825 // CHECK10-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
6826 // CHECK10-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6827 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6828 // CHECK10: omp.inner.for.inc:
6829 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6830 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6831 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
6832 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6833 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
6834 // CHECK10: omp.inner.for.end:
6835 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6836 // CHECK10: omp.loop.exit:
6837 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6838 // CHECK10-NEXT: ret void
6839 //
6840 //
6841 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
6842 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6843 // CHECK10-NEXT: entry:
6844 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6845 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6846 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6847 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6848 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6849 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6850 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6851 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6852 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6853 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6854 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6855 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6856 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6857 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6858 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6859 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6860 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
6861 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6862 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6863 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6864 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6865 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6866 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6867 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6868 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6869 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6870 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6871 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6872 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6873 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6874 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6875 // CHECK10: cond.true:
6876 // CHECK10-NEXT: br label [[COND_END:%.*]]
6877 // CHECK10: cond.false:
6878 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6879 // CHECK10-NEXT: br label [[COND_END]]
6880 // CHECK10: cond.end:
6881 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6882 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6883 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6884 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6885 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6886 // CHECK10: omp.inner.for.cond:
6887 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6888 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6889 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6890 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6891 // CHECK10: omp.inner.for.body:
6892 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6893 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6894 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6895 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6896 // CHECK10-NEXT: call void @_Z9gtid_testv()
6897 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6898 // CHECK10: omp.body.continue:
6899 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6900 // CHECK10: omp.inner.for.inc:
6901 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6902 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6903 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6904 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
6905 // CHECK10: omp.inner.for.end:
6906 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6907 // CHECK10: omp.loop.exit:
6908 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6909 // CHECK10-NEXT: ret void
6910 //
6911 //
6912 // CHECK10-LABEL: define {{[^@]+}}@main
6913 // CHECK10-SAME: () #[[ATTR3:[0-9]+]] {
6914 // CHECK10-NEXT: entry:
6915 // CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
6916 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6917 // CHECK10-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
6918 // CHECK10-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
6919 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
6920 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
6921 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
6922 // CHECK10-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
6923 // CHECK10-NEXT: store i32 0, i32* [[RETVAL]], align 4
6924 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6925 // CHECK10-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6926 // CHECK10-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
6927 // CHECK10-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6928 // CHECK10: omp_offload.failed:
6929 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
6930 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
6931 // CHECK10: omp_offload.cont:
6932 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6933 // CHECK10-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
6934 // CHECK10-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
6935 // CHECK10-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
6936 // CHECK10: omp_offload.failed2:
6937 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
6938 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT3]]
6939 // CHECK10: omp_offload.cont3:
6940 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
6941 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
6942 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
6943 // CHECK10-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
6944 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6945 // CHECK10-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
6946 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
6947 // CHECK10-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6948 // CHECK10-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
6949 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
6950 // CHECK10-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
6951 // CHECK10-NEXT: store i8* null, i8** [[TMP10]], align 8
6952 // CHECK10-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6953 // CHECK10-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6954 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6955 // CHECK10-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6956 // CHECK10-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6957 // CHECK10-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
6958 // CHECK10: omp_offload.failed5:
6959 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
6960 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT6]]
6961 // CHECK10: omp_offload.cont6:
6962 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
6963 // CHECK10-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
6964 // CHECK10-NEXT: ret i32 [[CALL]]
6965 //
6966 //
6967 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
6968 // CHECK10-SAME: () #[[ATTR1]] {
6969 // CHECK10-NEXT: entry:
6970 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
6971 // CHECK10-NEXT: ret void
6972 //
6973 //
6974 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
6975 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6976 // CHECK10-NEXT: entry:
6977 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6978 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6979 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6980 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6981 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6982 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6983 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6984 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6985 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6986 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6987 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6988 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6989 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6990 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6991 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6992 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6993 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6994 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6995 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6996 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6997 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6998 // CHECK10: cond.true:
6999 // CHECK10-NEXT: br label [[COND_END:%.*]]
7000 // CHECK10: cond.false:
7001 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7002 // CHECK10-NEXT: br label [[COND_END]]
7003 // CHECK10: cond.end:
7004 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7005 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7006 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7007 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7008 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7009 // CHECK10: omp.inner.for.cond:
7010 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7011 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7012 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7013 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7014 // CHECK10: omp.inner.for.body:
7015 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7016 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7017 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7018 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7019 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
7020 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7021 // CHECK10: omp.inner.for.inc:
7022 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7023 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7024 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
7025 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7026 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7027 // CHECK10: omp.inner.for.end:
7028 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7029 // CHECK10: omp.loop.exit:
7030 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7031 // CHECK10-NEXT: ret void
7032 //
7033 //
7034 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
7035 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7036 // CHECK10-NEXT: entry:
7037 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7038 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7039 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7040 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7041 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7042 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7043 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7044 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7045 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7046 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7047 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7048 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7049 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7050 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7051 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7052 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7053 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7054 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7055 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7056 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7057 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7058 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7059 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7060 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7061 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7062 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7063 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7064 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7065 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7066 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7067 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7068 // CHECK10: cond.true:
7069 // CHECK10-NEXT: br label [[COND_END:%.*]]
7070 // CHECK10: cond.false:
7071 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7072 // CHECK10-NEXT: br label [[COND_END]]
7073 // CHECK10: cond.end:
7074 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7075 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7076 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7077 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7078 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7079 // CHECK10: omp.inner.for.cond:
7080 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7081 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7082 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7083 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7084 // CHECK10: omp.inner.for.body:
7085 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7086 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7087 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7088 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7089 // CHECK10-NEXT: call void @_Z3fn4v()
7090 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7091 // CHECK10: omp.body.continue:
7092 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7093 // CHECK10: omp.inner.for.inc:
7094 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7095 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7096 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7097 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7098 // CHECK10: omp.inner.for.end:
7099 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7100 // CHECK10: omp.loop.exit:
7101 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7102 // CHECK10-NEXT: ret void
7103 //
7104 //
7105 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
7106 // CHECK10-SAME: () #[[ATTR1]] {
7107 // CHECK10-NEXT: entry:
7108 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
7109 // CHECK10-NEXT: ret void
7110 //
7111 //
7112 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
7113 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7114 // CHECK10-NEXT: entry:
7115 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7116 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7117 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7118 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7119 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7120 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7121 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7122 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7123 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7124 // CHECK10-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7125 // CHECK10-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
7126 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7127 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7128 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7129 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7130 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7131 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7132 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7133 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7134 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7135 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7136 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7137 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7138 // CHECK10: cond.true:
7139 // CHECK10-NEXT: br label [[COND_END:%.*]]
7140 // CHECK10: cond.false:
7141 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7142 // CHECK10-NEXT: br label [[COND_END]]
7143 // CHECK10: cond.end:
7144 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7145 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7146 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7147 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7148 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7149 // CHECK10: omp.inner.for.cond:
7150 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7151 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7152 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7153 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7154 // CHECK10: omp.inner.for.body:
7155 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7156 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7157 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7158 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7159 // CHECK10-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
7160 // CHECK10-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7161 // CHECK10-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
7162 // CHECK10-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
7163 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7164 // CHECK10: omp.inner.for.inc:
7165 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7166 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7167 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
7168 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7169 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7170 // CHECK10: omp.inner.for.end:
7171 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7172 // CHECK10: omp.loop.exit:
7173 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7174 // CHECK10-NEXT: ret void
7175 //
7176 //
7177 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..7
7178 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7179 // CHECK10-NEXT: entry:
7180 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7181 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7182 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7183 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7184 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7185 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7186 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7187 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7188 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7189 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7190 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7191 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7192 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7193 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7194 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7195 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7196 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7197 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7198 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7199 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7200 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7201 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7202 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7203 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7204 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7205 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7206 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7207 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7208 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7209 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7210 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7211 // CHECK10: cond.true:
7212 // CHECK10-NEXT: br label [[COND_END:%.*]]
7213 // CHECK10: cond.false:
7214 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7215 // CHECK10-NEXT: br label [[COND_END]]
7216 // CHECK10: cond.end:
7217 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7218 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7219 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7220 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7221 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7222 // CHECK10: omp.inner.for.cond:
7223 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7224 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7225 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7226 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7227 // CHECK10: omp.inner.for.body:
7228 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7229 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7230 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7231 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7232 // CHECK10-NEXT: call void @_Z3fn5v()
7233 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7234 // CHECK10: omp.body.continue:
7235 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7236 // CHECK10: omp.inner.for.inc:
7237 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7238 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7239 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7240 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7241 // CHECK10: omp.inner.for.end:
7242 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7243 // CHECK10: omp.loop.exit:
7244 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7245 // CHECK10-NEXT: ret void
7246 //
7247 //
7248 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
7249 // CHECK10-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
7250 // CHECK10-NEXT: entry:
7251 // CHECK10-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
7252 // CHECK10-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
7253 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
7254 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
7255 // CHECK10-NEXT: ret void
7256 //
7257 //
7258 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..8
7259 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
7260 // CHECK10-NEXT: entry:
7261 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7262 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7263 // CHECK10-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
7264 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7265 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7266 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7267 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7268 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7269 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7270 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7271 // CHECK10-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7272 // CHECK10-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
7273 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7274 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7275 // CHECK10-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
7276 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
7277 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7278 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7279 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7280 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7281 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7282 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7283 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7284 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7285 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
7286 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7287 // CHECK10: cond.true:
7288 // CHECK10-NEXT: br label [[COND_END:%.*]]
7289 // CHECK10: cond.false:
7290 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7291 // CHECK10-NEXT: br label [[COND_END]]
7292 // CHECK10: cond.end:
7293 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7294 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7295 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7296 // CHECK10-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7297 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7298 // CHECK10: omp.inner.for.cond:
7299 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7300 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7301 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7302 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7303 // CHECK10: omp.inner.for.body:
7304 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7305 // CHECK10-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
7306 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7307 // CHECK10-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
7308 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
7309 // CHECK10-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
7310 // CHECK10-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7311 // CHECK10: omp_if.then:
7312 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
7313 // CHECK10-NEXT: br label [[OMP_IF_END:%.*]]
7314 // CHECK10: omp_if.else:
7315 // CHECK10-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
7316 // CHECK10-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7317 // CHECK10-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
7318 // CHECK10-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
7319 // CHECK10-NEXT: br label [[OMP_IF_END]]
7320 // CHECK10: omp_if.end:
7321 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7322 // CHECK10: omp.inner.for.inc:
7323 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7324 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7325 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
7326 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7327 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7328 // CHECK10: omp.inner.for.end:
7329 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7330 // CHECK10: omp.loop.exit:
7331 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7332 // CHECK10-NEXT: ret void
7333 //
7334 //
7335 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..9
7336 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7337 // CHECK10-NEXT: entry:
7338 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7339 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7340 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7341 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7342 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7343 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7344 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7345 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7346 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7347 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7348 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7349 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7350 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7351 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7352 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7353 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7354 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7355 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7356 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7357 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7358 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7359 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7360 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7361 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7362 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7363 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7364 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7365 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7366 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7367 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7368 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7369 // CHECK10: cond.true:
7370 // CHECK10-NEXT: br label [[COND_END:%.*]]
7371 // CHECK10: cond.false:
7372 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7373 // CHECK10-NEXT: br label [[COND_END]]
7374 // CHECK10: cond.end:
7375 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7376 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7377 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7378 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7379 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7380 // CHECK10: omp.inner.for.cond:
7381 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7382 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7383 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7384 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7385 // CHECK10: omp.inner.for.body:
7386 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7387 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7388 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7389 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7390 // CHECK10-NEXT: call void @_Z3fn6v()
7391 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7392 // CHECK10: omp.body.continue:
7393 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7394 // CHECK10: omp.inner.for.inc:
7395 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7396 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7397 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7398 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7399 // CHECK10: omp.inner.for.end:
7400 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7401 // CHECK10: omp.loop.exit:
7402 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7403 // CHECK10-NEXT: ret void
7404 //
7405 //
7406 // CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
7407 // CHECK10-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
7408 // CHECK10-NEXT: entry:
7409 // CHECK10-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
7410 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7411 // CHECK10-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
7412 // CHECK10-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
7413 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
7414 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
7415 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
7416 // CHECK10-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
7417 // CHECK10-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
7418 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7419 // CHECK10-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
7420 // CHECK10-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
7421 // CHECK10-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7422 // CHECK10: omp_offload.failed:
7423 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
7424 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
7425 // CHECK10: omp_offload.cont:
7426 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7427 // CHECK10-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
7428 // CHECK10-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
7429 // CHECK10-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
7430 // CHECK10: omp_offload.failed2:
7431 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
7432 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT3]]
7433 // CHECK10: omp_offload.cont3:
7434 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
7435 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
7436 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
7437 // CHECK10-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
7438 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7439 // CHECK10-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
7440 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
7441 // CHECK10-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7442 // CHECK10-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
7443 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
7444 // CHECK10-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7445 // CHECK10-NEXT: store i8* null, i8** [[TMP10]], align 8
7446 // CHECK10-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7447 // CHECK10-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7448 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7449 // CHECK10-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7450 // CHECK10-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7451 // CHECK10-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
7452 // CHECK10: omp_offload.failed5:
7453 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
7454 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT6]]
7455 // CHECK10: omp_offload.cont6:
7456 // CHECK10-NEXT: ret i32 0
7457 //
7458 //
7459 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
7460 // CHECK10-SAME: () #[[ATTR1]] {
7461 // CHECK10-NEXT: entry:
7462 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
7463 // CHECK10-NEXT: ret void
7464 //
7465 //
7466 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..10
7467 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7468 // CHECK10-NEXT: entry:
7469 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7470 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7471 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7472 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7473 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7474 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7475 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7476 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7477 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7478 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7479 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7480 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7481 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7482 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7483 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7484 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7485 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7486 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7487 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7488 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7489 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7490 // CHECK10: cond.true:
7491 // CHECK10-NEXT: br label [[COND_END:%.*]]
7492 // CHECK10: cond.false:
7493 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7494 // CHECK10-NEXT: br label [[COND_END]]
7495 // CHECK10: cond.end:
7496 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7497 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7498 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7499 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7500 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7501 // CHECK10: omp.inner.for.cond:
7502 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7503 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7504 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7505 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7506 // CHECK10: omp.inner.for.body:
7507 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7508 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7509 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7510 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7511 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
7512 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7513 // CHECK10: omp.inner.for.inc:
7514 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7515 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7516 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
7517 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7518 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7519 // CHECK10: omp.inner.for.end:
7520 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7521 // CHECK10: omp.loop.exit:
7522 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7523 // CHECK10-NEXT: ret void
7524 //
7525 //
7526 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..11
7527 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7528 // CHECK10-NEXT: entry:
7529 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7530 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7531 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7532 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7533 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7534 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7535 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7536 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7537 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7538 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7539 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7540 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7541 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7542 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7543 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7544 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7545 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7546 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7547 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7548 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7549 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7550 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7551 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7552 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7553 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7554 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7555 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7556 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7557 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7558 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7559 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7560 // CHECK10: cond.true:
7561 // CHECK10-NEXT: br label [[COND_END:%.*]]
7562 // CHECK10: cond.false:
7563 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7564 // CHECK10-NEXT: br label [[COND_END]]
7565 // CHECK10: cond.end:
7566 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7567 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7568 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7569 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7570 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7571 // CHECK10: omp.inner.for.cond:
7572 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7573 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7574 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7575 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7576 // CHECK10: omp.inner.for.body:
7577 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7578 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7579 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7580 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7581 // CHECK10-NEXT: call void @_Z3fn1v()
7582 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7583 // CHECK10: omp.body.continue:
7584 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7585 // CHECK10: omp.inner.for.inc:
7586 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7587 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7588 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7589 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7590 // CHECK10: omp.inner.for.end:
7591 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7592 // CHECK10: omp.loop.exit:
7593 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7594 // CHECK10-NEXT: ret void
7595 //
7596 //
7597 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
7598 // CHECK10-SAME: () #[[ATTR1]] {
7599 // CHECK10-NEXT: entry:
7600 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
7601 // CHECK10-NEXT: ret void
7602 //
7603 //
7604 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..12
7605 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7606 // CHECK10-NEXT: entry:
7607 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7608 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7609 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7610 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7611 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7612 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7613 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7614 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7615 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7616 // CHECK10-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7617 // CHECK10-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
7618 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7619 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7620 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7621 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7622 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7623 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7624 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7625 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7626 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7627 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7628 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7629 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7630 // CHECK10: cond.true:
7631 // CHECK10-NEXT: br label [[COND_END:%.*]]
7632 // CHECK10: cond.false:
7633 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7634 // CHECK10-NEXT: br label [[COND_END]]
7635 // CHECK10: cond.end:
7636 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7637 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7638 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7639 // CHECK10-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7640 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7641 // CHECK10: omp.inner.for.cond:
7642 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7643 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7644 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7645 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7646 // CHECK10: omp.inner.for.body:
7647 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7648 // CHECK10-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7649 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7650 // CHECK10-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7651 // CHECK10-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
7652 // CHECK10-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7653 // CHECK10-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
7654 // CHECK10-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
7655 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7656 // CHECK10: omp.inner.for.inc:
7657 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7658 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7659 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
7660 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7661 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7662 // CHECK10: omp.inner.for.end:
7663 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7664 // CHECK10: omp.loop.exit:
7665 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7666 // CHECK10-NEXT: ret void
7667 //
7668 //
7669 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..13
7670 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7671 // CHECK10-NEXT: entry:
7672 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7673 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7674 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7675 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7676 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7677 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7678 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7679 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7680 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7681 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7682 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7683 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7684 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7685 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7686 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7687 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7688 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7689 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7690 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7691 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7692 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7693 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7694 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7695 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7696 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7697 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7698 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7699 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7700 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7701 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7702 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7703 // CHECK10: cond.true:
7704 // CHECK10-NEXT: br label [[COND_END:%.*]]
7705 // CHECK10: cond.false:
7706 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7707 // CHECK10-NEXT: br label [[COND_END]]
7708 // CHECK10: cond.end:
7709 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7710 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7711 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7712 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7713 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7714 // CHECK10: omp.inner.for.cond:
7715 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7716 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7717 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7718 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7719 // CHECK10: omp.inner.for.body:
7720 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7721 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7722 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7723 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7724 // CHECK10-NEXT: call void @_Z3fn2v()
7725 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7726 // CHECK10: omp.body.continue:
7727 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7728 // CHECK10: omp.inner.for.inc:
7729 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7730 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7731 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7732 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7733 // CHECK10: omp.inner.for.end:
7734 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7735 // CHECK10: omp.loop.exit:
7736 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7737 // CHECK10-NEXT: ret void
7738 //
7739 //
7740 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
7741 // CHECK10-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
7742 // CHECK10-NEXT: entry:
7743 // CHECK10-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
7744 // CHECK10-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
7745 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
7746 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
7747 // CHECK10-NEXT: ret void
7748 //
7749 //
7750 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..14
7751 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
7752 // CHECK10-NEXT: entry:
7753 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7754 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7755 // CHECK10-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
7756 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7757 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7758 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7759 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7760 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7761 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7762 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7763 // CHECK10-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7764 // CHECK10-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
7765 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7766 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7767 // CHECK10-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
7768 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
7769 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7770 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7771 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7772 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7773 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7774 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7775 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7776 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7777 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
7778 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7779 // CHECK10: cond.true:
7780 // CHECK10-NEXT: br label [[COND_END:%.*]]
7781 // CHECK10: cond.false:
7782 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7783 // CHECK10-NEXT: br label [[COND_END]]
7784 // CHECK10: cond.end:
7785 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7786 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7787 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7788 // CHECK10-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7789 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7790 // CHECK10: omp.inner.for.cond:
7791 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7792 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7793 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7794 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7795 // CHECK10: omp.inner.for.body:
7796 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7797 // CHECK10-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
7798 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7799 // CHECK10-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
7800 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
7801 // CHECK10-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
7802 // CHECK10-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7803 // CHECK10: omp_if.then:
7804 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
7805 // CHECK10-NEXT: br label [[OMP_IF_END:%.*]]
7806 // CHECK10: omp_if.else:
7807 // CHECK10-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
7808 // CHECK10-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7809 // CHECK10-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
7810 // CHECK10-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
7811 // CHECK10-NEXT: br label [[OMP_IF_END]]
7812 // CHECK10: omp_if.end:
7813 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7814 // CHECK10: omp.inner.for.inc:
7815 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7816 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7817 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
7818 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7819 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7820 // CHECK10: omp.inner.for.end:
7821 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7822 // CHECK10: omp.loop.exit:
7823 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7824 // CHECK10-NEXT: ret void
7825 //
7826 //
7827 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..15
7828 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7829 // CHECK10-NEXT: entry:
7830 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7831 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7832 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7833 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7834 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7835 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
7836 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7837 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7838 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7839 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7840 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
7841 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7842 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7843 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7844 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7845 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7846 // CHECK10-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
7847 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7848 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7849 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7850 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7851 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7852 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7853 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7854 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7855 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7856 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7857 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7858 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7859 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7860 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7861 // CHECK10: cond.true:
7862 // CHECK10-NEXT: br label [[COND_END:%.*]]
7863 // CHECK10: cond.false:
7864 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7865 // CHECK10-NEXT: br label [[COND_END]]
7866 // CHECK10: cond.end:
7867 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7868 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7869 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7870 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7871 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7872 // CHECK10: omp.inner.for.cond:
7873 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7874 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7875 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7876 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7877 // CHECK10: omp.inner.for.body:
7878 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7879 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7880 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7881 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7882 // CHECK10-NEXT: call void @_Z3fn3v()
7883 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7884 // CHECK10: omp.body.continue:
7885 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7886 // CHECK10: omp.inner.for.inc:
7887 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7888 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7889 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7890 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
7891 // CHECK10: omp.inner.for.end:
7892 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7893 // CHECK10: omp.loop.exit:
7894 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7895 // CHECK10-NEXT: ret void
7896 //
7897 //
7898 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7899 // CHECK10-SAME: () #[[ATTR5:[0-9]+]] {
7900 // CHECK10-NEXT: entry:
7901 // CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
7902 // CHECK10-NEXT: ret void
7903 //
7904 //
7905 // CHECK13-LABEL: define {{[^@]+}}@_Z9gtid_testv
7906 // CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
7907 // CHECK13-NEXT: entry:
7908 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
7909 // CHECK13-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
7910 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
7911 // CHECK13-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
7912 // CHECK13-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
7913 // CHECK13-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7914 // CHECK13: omp_offload.failed:
7915 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
7916 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT]]
7917 // CHECK13: omp_offload.cont:
7918 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7919 // CHECK13-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
7920 // CHECK13-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
7921 // CHECK13-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
7922 // CHECK13: omp_offload.failed2:
7923 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
7924 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT3]]
7925 // CHECK13: omp_offload.cont3:
7926 // CHECK13-NEXT: ret void
7927 //
7928 //
7929 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
7930 // CHECK13-SAME: () #[[ATTR1:[0-9]+]] {
7931 // CHECK13-NEXT: entry:
7932 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
7933 // CHECK13-NEXT: ret void
7934 //
7935 //
7936 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined.
7937 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7938 // CHECK13-NEXT: entry:
7939 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7940 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7941 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7942 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
7943 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7944 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7945 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7946 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7947 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
7948 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7949 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7950 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7951 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7952 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7953 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7954 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7955 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7956 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7957 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7958 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7959 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7960 // CHECK13: cond.true:
7961 // CHECK13-NEXT: br label [[COND_END:%.*]]
7962 // CHECK13: cond.false:
7963 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7964 // CHECK13-NEXT: br label [[COND_END]]
7965 // CHECK13: cond.end:
7966 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7967 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7968 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7969 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7970 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7971 // CHECK13: omp.inner.for.cond:
7972 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7973 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7974 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7975 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7976 // CHECK13: omp.inner.for.body:
7977 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7978 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7979 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7980 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7981 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
7982 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7983 // CHECK13: omp.inner.for.inc:
7984 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7985 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7986 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
7987 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7988 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
7989 // CHECK13: omp.inner.for.end:
7990 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7991 // CHECK13: omp.loop.exit:
7992 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7993 // CHECK13-NEXT: ret void
7994 //
7995 //
7996 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
7997 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7998 // CHECK13-NEXT: entry:
7999 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8000 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8001 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8002 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8003 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8004 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8005 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8006 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8007 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8008 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8009 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8010 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8011 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8012 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8013 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8014 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8015 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8016 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8017 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8018 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8019 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8020 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8021 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8022 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8023 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8024 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8025 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8026 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8027 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8028 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8029 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8030 // CHECK13: cond.true:
8031 // CHECK13-NEXT: br label [[COND_END:%.*]]
8032 // CHECK13: cond.false:
8033 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8034 // CHECK13-NEXT: br label [[COND_END]]
8035 // CHECK13: cond.end:
8036 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8037 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8038 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8039 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8040 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8041 // CHECK13: omp.inner.for.cond:
8042 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8043 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8044 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8045 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8046 // CHECK13: omp.inner.for.body:
8047 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8048 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8049 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8050 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8051 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8052 // CHECK13: omp.body.continue:
8053 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8054 // CHECK13: omp.inner.for.inc:
8055 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8056 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8057 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8058 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8059 // CHECK13: omp.inner.for.end:
8060 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8061 // CHECK13: omp.loop.exit:
8062 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8063 // CHECK13-NEXT: ret void
8064 //
8065 //
8066 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
8067 // CHECK13-SAME: () #[[ATTR1]] {
8068 // CHECK13-NEXT: entry:
8069 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
8070 // CHECK13-NEXT: ret void
8071 //
8072 //
8073 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..2
8074 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8075 // CHECK13-NEXT: entry:
8076 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8077 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8078 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8079 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8080 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8081 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8082 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8083 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8084 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8085 // CHECK13-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8086 // CHECK13-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
8087 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8088 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8089 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8090 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8091 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8092 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8093 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8094 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8095 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8096 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8097 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8098 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8099 // CHECK13: cond.true:
8100 // CHECK13-NEXT: br label [[COND_END:%.*]]
8101 // CHECK13: cond.false:
8102 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8103 // CHECK13-NEXT: br label [[COND_END]]
8104 // CHECK13: cond.end:
8105 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8106 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8107 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8108 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8109 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8110 // CHECK13: omp.inner.for.cond:
8111 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8112 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8113 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8114 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8115 // CHECK13: omp.inner.for.body:
8116 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8117 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8118 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8119 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8120 // CHECK13-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8121 // CHECK13-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8122 // CHECK13-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
8123 // CHECK13-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8124 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8125 // CHECK13: omp.inner.for.inc:
8126 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8127 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8128 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
8129 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8130 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8131 // CHECK13: omp.inner.for.end:
8132 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8133 // CHECK13: omp.loop.exit:
8134 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8135 // CHECK13-NEXT: ret void
8136 //
8137 //
8138 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
8139 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8140 // CHECK13-NEXT: entry:
8141 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8142 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8143 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8144 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8145 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8146 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8147 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8148 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8149 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8150 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8151 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8152 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8153 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8154 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8155 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8156 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8157 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8158 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8159 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8160 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8161 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8162 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8163 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8164 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8165 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8166 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8167 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8168 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8169 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8170 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8171 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8172 // CHECK13: cond.true:
8173 // CHECK13-NEXT: br label [[COND_END:%.*]]
8174 // CHECK13: cond.false:
8175 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8176 // CHECK13-NEXT: br label [[COND_END]]
8177 // CHECK13: cond.end:
8178 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8179 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8180 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8181 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8182 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8183 // CHECK13: omp.inner.for.cond:
8184 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8185 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8186 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8187 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8188 // CHECK13: omp.inner.for.body:
8189 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8190 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8191 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8192 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8193 // CHECK13-NEXT: call void @_Z9gtid_testv()
8194 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8195 // CHECK13: omp.body.continue:
8196 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8197 // CHECK13: omp.inner.for.inc:
8198 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8199 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8200 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8201 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8202 // CHECK13: omp.inner.for.end:
8203 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8204 // CHECK13: omp.loop.exit:
8205 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8206 // CHECK13-NEXT: ret void
8207 //
8208 //
8209 // CHECK13-LABEL: define {{[^@]+}}@main
8210 // CHECK13-SAME: () #[[ATTR3:[0-9]+]] {
8211 // CHECK13-NEXT: entry:
8212 // CHECK13-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
8213 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8214 // CHECK13-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
8215 // CHECK13-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
8216 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
8217 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
8218 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
8219 // CHECK13-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
8220 // CHECK13-NEXT: store i32 0, i32* [[RETVAL]], align 4
8221 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8222 // CHECK13-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
8223 // CHECK13-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
8224 // CHECK13-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8225 // CHECK13: omp_offload.failed:
8226 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
8227 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT]]
8228 // CHECK13: omp_offload.cont:
8229 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8230 // CHECK13-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
8231 // CHECK13-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
8232 // CHECK13-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
8233 // CHECK13: omp_offload.failed2:
8234 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
8235 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT3]]
8236 // CHECK13: omp_offload.cont3:
8237 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
8238 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
8239 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
8240 // CHECK13-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
8241 // CHECK13-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8242 // CHECK13-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
8243 // CHECK13-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
8244 // CHECK13-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8245 // CHECK13-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
8246 // CHECK13-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
8247 // CHECK13-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8248 // CHECK13-NEXT: store i8* null, i8** [[TMP10]], align 8
8249 // CHECK13-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8250 // CHECK13-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8251 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8252 // CHECK13-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8253 // CHECK13-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
8254 // CHECK13-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
8255 // CHECK13: omp_offload.failed5:
8256 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
8257 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT6]]
8258 // CHECK13: omp_offload.cont6:
8259 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
8260 // CHECK13-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
8261 // CHECK13-NEXT: ret i32 [[CALL]]
8262 //
8263 //
8264 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
8265 // CHECK13-SAME: () #[[ATTR1]] {
8266 // CHECK13-NEXT: entry:
8267 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
8268 // CHECK13-NEXT: ret void
8269 //
8270 //
8271 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..4
8272 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8273 // CHECK13-NEXT: entry:
8274 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8275 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8276 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8277 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8278 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8279 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8280 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8281 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8282 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8283 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8284 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8285 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8286 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8287 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8288 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8289 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8290 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8291 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8292 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8293 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8294 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8295 // CHECK13: cond.true:
8296 // CHECK13-NEXT: br label [[COND_END:%.*]]
8297 // CHECK13: cond.false:
8298 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8299 // CHECK13-NEXT: br label [[COND_END]]
8300 // CHECK13: cond.end:
8301 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8302 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8303 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8304 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8305 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8306 // CHECK13: omp.inner.for.cond:
8307 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8308 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8309 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8310 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8311 // CHECK13: omp.inner.for.body:
8312 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8313 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8314 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8315 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8316 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
8317 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8318 // CHECK13: omp.inner.for.inc:
8319 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8320 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8321 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
8322 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8323 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8324 // CHECK13: omp.inner.for.end:
8325 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8326 // CHECK13: omp.loop.exit:
8327 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8328 // CHECK13-NEXT: ret void
8329 //
8330 //
8331 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..5
8332 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8333 // CHECK13-NEXT: entry:
8334 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8335 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8336 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8337 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8338 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8339 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8340 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8341 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8342 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8343 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8344 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8345 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8346 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8347 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8348 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8349 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8350 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8351 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8352 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8353 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8354 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8355 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8356 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8357 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8358 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8359 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8360 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8361 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8362 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8363 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8364 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8365 // CHECK13: cond.true:
8366 // CHECK13-NEXT: br label [[COND_END:%.*]]
8367 // CHECK13: cond.false:
8368 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8369 // CHECK13-NEXT: br label [[COND_END]]
8370 // CHECK13: cond.end:
8371 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8372 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8373 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8374 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8375 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8376 // CHECK13: omp.inner.for.cond:
8377 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8378 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8379 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8380 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8381 // CHECK13: omp.inner.for.body:
8382 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8383 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8384 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8385 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8386 // CHECK13-NEXT: call void @_Z3fn4v()
8387 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8388 // CHECK13: omp.body.continue:
8389 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8390 // CHECK13: omp.inner.for.inc:
8391 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8392 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8393 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8394 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8395 // CHECK13: omp.inner.for.end:
8396 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8397 // CHECK13: omp.loop.exit:
8398 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8399 // CHECK13-NEXT: ret void
8400 //
8401 //
8402 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
8403 // CHECK13-SAME: () #[[ATTR1]] {
8404 // CHECK13-NEXT: entry:
8405 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
8406 // CHECK13-NEXT: ret void
8407 //
8408 //
8409 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..6
8410 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8411 // CHECK13-NEXT: entry:
8412 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8413 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8414 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8415 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8416 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8417 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8418 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8419 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8420 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8421 // CHECK13-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8422 // CHECK13-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
8423 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8424 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8425 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8426 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8427 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8428 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8429 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8430 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8431 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8432 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8433 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8434 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8435 // CHECK13: cond.true:
8436 // CHECK13-NEXT: br label [[COND_END:%.*]]
8437 // CHECK13: cond.false:
8438 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8439 // CHECK13-NEXT: br label [[COND_END]]
8440 // CHECK13: cond.end:
8441 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8442 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8443 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8444 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8445 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8446 // CHECK13: omp.inner.for.cond:
8447 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8448 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8449 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8450 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8451 // CHECK13: omp.inner.for.body:
8452 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8453 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8454 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8455 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8456 // CHECK13-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8457 // CHECK13-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8458 // CHECK13-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
8459 // CHECK13-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8460 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8461 // CHECK13: omp.inner.for.inc:
8462 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8463 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8464 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
8465 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8466 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8467 // CHECK13: omp.inner.for.end:
8468 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8469 // CHECK13: omp.loop.exit:
8470 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8471 // CHECK13-NEXT: ret void
8472 //
8473 //
8474 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..7
8475 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8476 // CHECK13-NEXT: entry:
8477 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8478 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8479 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8480 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8481 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8482 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8483 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8484 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8485 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8486 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8487 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8488 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8489 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8490 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8491 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8492 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8493 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8494 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8495 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8496 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8497 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8498 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8499 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8500 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8501 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8502 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8503 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8504 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8505 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8506 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8507 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8508 // CHECK13: cond.true:
8509 // CHECK13-NEXT: br label [[COND_END:%.*]]
8510 // CHECK13: cond.false:
8511 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8512 // CHECK13-NEXT: br label [[COND_END]]
8513 // CHECK13: cond.end:
8514 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8515 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8516 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8517 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8518 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8519 // CHECK13: omp.inner.for.cond:
8520 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8521 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8522 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8523 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8524 // CHECK13: omp.inner.for.body:
8525 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8526 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8527 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8528 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8529 // CHECK13-NEXT: call void @_Z3fn5v()
8530 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8531 // CHECK13: omp.body.continue:
8532 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8533 // CHECK13: omp.inner.for.inc:
8534 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8535 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8536 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8537 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8538 // CHECK13: omp.inner.for.end:
8539 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8540 // CHECK13: omp.loop.exit:
8541 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8542 // CHECK13-NEXT: ret void
8543 //
8544 //
8545 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
8546 // CHECK13-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
8547 // CHECK13-NEXT: entry:
8548 // CHECK13-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
8549 // CHECK13-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
8550 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
8551 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
8552 // CHECK13-NEXT: ret void
8553 //
8554 //
8555 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..8
8556 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
8557 // CHECK13-NEXT: entry:
8558 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8559 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8560 // CHECK13-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
8561 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8562 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8563 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8564 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8565 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8566 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8567 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8568 // CHECK13-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8569 // CHECK13-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
8570 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8571 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8572 // CHECK13-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
8573 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
8574 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8575 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8576 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8577 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8578 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8579 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8580 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8581 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8582 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
8583 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8584 // CHECK13: cond.true:
8585 // CHECK13-NEXT: br label [[COND_END:%.*]]
8586 // CHECK13: cond.false:
8587 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8588 // CHECK13-NEXT: br label [[COND_END]]
8589 // CHECK13: cond.end:
8590 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8591 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8592 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8593 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
8594 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8595 // CHECK13: omp.inner.for.cond:
8596 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8597 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8598 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8599 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8600 // CHECK13: omp.inner.for.body:
8601 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8602 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
8603 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8604 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
8605 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
8606 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
8607 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8608 // CHECK13: omp_if.then:
8609 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
8610 // CHECK13-NEXT: br label [[OMP_IF_END:%.*]]
8611 // CHECK13: omp_if.else:
8612 // CHECK13-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
8613 // CHECK13-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8614 // CHECK13-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
8615 // CHECK13-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
8616 // CHECK13-NEXT: br label [[OMP_IF_END]]
8617 // CHECK13: omp_if.end:
8618 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8619 // CHECK13: omp.inner.for.inc:
8620 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8621 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8622 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
8623 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8624 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8625 // CHECK13: omp.inner.for.end:
8626 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8627 // CHECK13: omp.loop.exit:
8628 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
8629 // CHECK13-NEXT: ret void
8630 //
8631 //
8632 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..9
8633 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8634 // CHECK13-NEXT: entry:
8635 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8636 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8637 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8638 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8639 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8640 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8641 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8642 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8643 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8644 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8645 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8646 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8647 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8648 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8649 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8650 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8651 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8652 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8653 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8654 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8655 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8656 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8657 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8658 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8659 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8660 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8661 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8662 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8663 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8664 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8665 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8666 // CHECK13: cond.true:
8667 // CHECK13-NEXT: br label [[COND_END:%.*]]
8668 // CHECK13: cond.false:
8669 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8670 // CHECK13-NEXT: br label [[COND_END]]
8671 // CHECK13: cond.end:
8672 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8673 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8674 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8675 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8676 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8677 // CHECK13: omp.inner.for.cond:
8678 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8679 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8680 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8681 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8682 // CHECK13: omp.inner.for.body:
8683 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8684 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8685 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8686 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8687 // CHECK13-NEXT: call void @_Z3fn6v()
8688 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8689 // CHECK13: omp.body.continue:
8690 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8691 // CHECK13: omp.inner.for.inc:
8692 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8693 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8694 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8695 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8696 // CHECK13: omp.inner.for.end:
8697 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8698 // CHECK13: omp.loop.exit:
8699 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8700 // CHECK13-NEXT: ret void
8701 //
8702 //
8703 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
8704 // CHECK13-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
8705 // CHECK13-NEXT: entry:
8706 // CHECK13-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
8707 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8708 // CHECK13-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
8709 // CHECK13-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
8710 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
8711 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
8712 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
8713 // CHECK13-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
8714 // CHECK13-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
8715 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8716 // CHECK13-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
8717 // CHECK13-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
8718 // CHECK13-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8719 // CHECK13: omp_offload.failed:
8720 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
8721 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT]]
8722 // CHECK13: omp_offload.cont:
8723 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8724 // CHECK13-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
8725 // CHECK13-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
8726 // CHECK13-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
8727 // CHECK13: omp_offload.failed2:
8728 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
8729 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT3]]
8730 // CHECK13: omp_offload.cont3:
8731 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
8732 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
8733 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
8734 // CHECK13-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
8735 // CHECK13-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8736 // CHECK13-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
8737 // CHECK13-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
8738 // CHECK13-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8739 // CHECK13-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
8740 // CHECK13-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
8741 // CHECK13-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8742 // CHECK13-NEXT: store i8* null, i8** [[TMP10]], align 8
8743 // CHECK13-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8744 // CHECK13-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8745 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8746 // CHECK13-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8747 // CHECK13-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
8748 // CHECK13-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
8749 // CHECK13: omp_offload.failed5:
8750 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
8751 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT6]]
8752 // CHECK13: omp_offload.cont6:
8753 // CHECK13-NEXT: ret i32 0
8754 //
8755 //
8756 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
8757 // CHECK13-SAME: () #[[ATTR1]] {
8758 // CHECK13-NEXT: entry:
8759 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
8760 // CHECK13-NEXT: ret void
8761 //
8762 //
8763 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..10
8764 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8765 // CHECK13-NEXT: entry:
8766 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8767 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8768 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8769 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8770 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8771 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8772 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8773 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8774 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8775 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8776 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8777 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8778 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8779 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8780 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8781 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8782 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8783 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8784 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8785 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8786 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8787 // CHECK13: cond.true:
8788 // CHECK13-NEXT: br label [[COND_END:%.*]]
8789 // CHECK13: cond.false:
8790 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8791 // CHECK13-NEXT: br label [[COND_END]]
8792 // CHECK13: cond.end:
8793 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8794 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8795 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8796 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8797 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8798 // CHECK13: omp.inner.for.cond:
8799 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8800 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8801 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8802 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8803 // CHECK13: omp.inner.for.body:
8804 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8805 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8806 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8807 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8808 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
8809 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8810 // CHECK13: omp.inner.for.inc:
8811 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8812 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8813 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
8814 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8815 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8816 // CHECK13: omp.inner.for.end:
8817 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8818 // CHECK13: omp.loop.exit:
8819 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8820 // CHECK13-NEXT: ret void
8821 //
8822 //
8823 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..11
8824 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8825 // CHECK13-NEXT: entry:
8826 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8827 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8828 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8829 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8830 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8831 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8832 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8833 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8834 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8835 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8836 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8837 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8838 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8839 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8840 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8841 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8842 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8843 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8844 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8845 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8846 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8847 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8848 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8849 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8850 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8851 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8852 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8853 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8854 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8855 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8856 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8857 // CHECK13: cond.true:
8858 // CHECK13-NEXT: br label [[COND_END:%.*]]
8859 // CHECK13: cond.false:
8860 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8861 // CHECK13-NEXT: br label [[COND_END]]
8862 // CHECK13: cond.end:
8863 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8864 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8865 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8866 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8867 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8868 // CHECK13: omp.inner.for.cond:
8869 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8870 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8871 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8872 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8873 // CHECK13: omp.inner.for.body:
8874 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8875 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8876 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8877 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
8878 // CHECK13-NEXT: call void @_Z3fn1v()
8879 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8880 // CHECK13: omp.body.continue:
8881 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8882 // CHECK13: omp.inner.for.inc:
8883 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8884 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8885 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
8886 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8887 // CHECK13: omp.inner.for.end:
8888 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8889 // CHECK13: omp.loop.exit:
8890 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8891 // CHECK13-NEXT: ret void
8892 //
8893 //
8894 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
8895 // CHECK13-SAME: () #[[ATTR1]] {
8896 // CHECK13-NEXT: entry:
8897 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
8898 // CHECK13-NEXT: ret void
8899 //
8900 //
8901 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..12
8902 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8903 // CHECK13-NEXT: entry:
8904 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8905 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8906 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8907 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8908 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8909 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8910 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8911 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8912 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8913 // CHECK13-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8914 // CHECK13-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
8915 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8916 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8917 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8918 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8919 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8920 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8921 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8922 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8923 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8924 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8925 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8926 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8927 // CHECK13: cond.true:
8928 // CHECK13-NEXT: br label [[COND_END:%.*]]
8929 // CHECK13: cond.false:
8930 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8931 // CHECK13-NEXT: br label [[COND_END]]
8932 // CHECK13: cond.end:
8933 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8934 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8935 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8936 // CHECK13-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8937 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8938 // CHECK13: omp.inner.for.cond:
8939 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8940 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8941 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8942 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8943 // CHECK13: omp.inner.for.body:
8944 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8945 // CHECK13-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8946 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8947 // CHECK13-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8948 // CHECK13-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8949 // CHECK13-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8950 // CHECK13-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
8951 // CHECK13-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
8952 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8953 // CHECK13: omp.inner.for.inc:
8954 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8955 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8956 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
8957 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8958 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8959 // CHECK13: omp.inner.for.end:
8960 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8961 // CHECK13: omp.loop.exit:
8962 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8963 // CHECK13-NEXT: ret void
8964 //
8965 //
8966 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..13
8967 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8968 // CHECK13-NEXT: entry:
8969 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8970 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8971 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8972 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8973 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8974 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8975 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8976 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8977 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8978 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8979 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8980 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8981 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8982 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8983 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8984 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8985 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
8986 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8987 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8988 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8989 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8990 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8991 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8992 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8993 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8994 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8995 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8996 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8997 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8998 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8999 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9000 // CHECK13: cond.true:
9001 // CHECK13-NEXT: br label [[COND_END:%.*]]
9002 // CHECK13: cond.false:
9003 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9004 // CHECK13-NEXT: br label [[COND_END]]
9005 // CHECK13: cond.end:
9006 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9007 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9008 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9009 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9010 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9011 // CHECK13: omp.inner.for.cond:
9012 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9013 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9014 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9015 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9016 // CHECK13: omp.inner.for.body:
9017 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9018 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9019 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9020 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9021 // CHECK13-NEXT: call void @_Z3fn2v()
9022 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9023 // CHECK13: omp.body.continue:
9024 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9025 // CHECK13: omp.inner.for.inc:
9026 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9027 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9028 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9029 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9030 // CHECK13: omp.inner.for.end:
9031 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9032 // CHECK13: omp.loop.exit:
9033 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9034 // CHECK13-NEXT: ret void
9035 //
9036 //
9037 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
9038 // CHECK13-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
9039 // CHECK13-NEXT: entry:
9040 // CHECK13-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
9041 // CHECK13-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
9042 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
9043 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
9044 // CHECK13-NEXT: ret void
9045 //
9046 //
9047 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..14
9048 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
9049 // CHECK13-NEXT: entry:
9050 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9051 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9052 // CHECK13-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
9053 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9054 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9055 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9056 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9057 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9058 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9059 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9060 // CHECK13-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9061 // CHECK13-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
9062 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9063 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9064 // CHECK13-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
9065 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
9066 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9067 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9068 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9069 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9070 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9071 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9072 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9073 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9074 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
9075 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9076 // CHECK13: cond.true:
9077 // CHECK13-NEXT: br label [[COND_END:%.*]]
9078 // CHECK13: cond.false:
9079 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9080 // CHECK13-NEXT: br label [[COND_END]]
9081 // CHECK13: cond.end:
9082 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9083 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9084 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9085 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9086 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9087 // CHECK13: omp.inner.for.cond:
9088 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9089 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9090 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9091 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9092 // CHECK13: omp.inner.for.body:
9093 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9094 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9095 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9096 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9097 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
9098 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
9099 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9100 // CHECK13: omp_if.then:
9101 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
9102 // CHECK13-NEXT: br label [[OMP_IF_END:%.*]]
9103 // CHECK13: omp_if.else:
9104 // CHECK13-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
9105 // CHECK13-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9106 // CHECK13-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
9107 // CHECK13-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
9108 // CHECK13-NEXT: br label [[OMP_IF_END]]
9109 // CHECK13: omp_if.end:
9110 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9111 // CHECK13: omp.inner.for.inc:
9112 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9113 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9114 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9115 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9116 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9117 // CHECK13: omp.inner.for.end:
9118 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9119 // CHECK13: omp.loop.exit:
9120 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9121 // CHECK13-NEXT: ret void
9122 //
9123 //
9124 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..15
9125 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9126 // CHECK13-NEXT: entry:
9127 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9128 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9129 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9130 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9131 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9132 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9133 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9134 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9135 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9136 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9137 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9138 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9139 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9140 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9141 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9142 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9143 // CHECK13-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9144 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9145 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9146 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9147 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9148 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9149 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9150 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9151 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9152 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9153 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9154 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9155 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9156 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9157 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9158 // CHECK13: cond.true:
9159 // CHECK13-NEXT: br label [[COND_END:%.*]]
9160 // CHECK13: cond.false:
9161 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9162 // CHECK13-NEXT: br label [[COND_END]]
9163 // CHECK13: cond.end:
9164 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9165 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9166 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9167 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9168 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9169 // CHECK13: omp.inner.for.cond:
9170 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9171 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9172 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9173 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9174 // CHECK13: omp.inner.for.body:
9175 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9176 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9177 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9178 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9179 // CHECK13-NEXT: call void @_Z3fn3v()
9180 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9181 // CHECK13: omp.body.continue:
9182 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9183 // CHECK13: omp.inner.for.inc:
9184 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9185 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9186 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9187 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9188 // CHECK13: omp.inner.for.end:
9189 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9190 // CHECK13: omp.loop.exit:
9191 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9192 // CHECK13-NEXT: ret void
9193 //
9194 //
9195 // CHECK13-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9196 // CHECK13-SAME: () #[[ATTR5:[0-9]+]] {
9197 // CHECK13-NEXT: entry:
9198 // CHECK13-NEXT: call void @__tgt_register_requires(i64 1)
9199 // CHECK13-NEXT: ret void
9200 //
9201 //
9202 // CHECK14-LABEL: define {{[^@]+}}@_Z9gtid_testv
9203 // CHECK14-SAME: () #[[ATTR0:[0-9]+]] {
9204 // CHECK14-NEXT: entry:
9205 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9206 // CHECK14-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
9207 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
9208 // CHECK14-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
9209 // CHECK14-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
9210 // CHECK14-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9211 // CHECK14: omp_offload.failed:
9212 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
9213 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT]]
9214 // CHECK14: omp_offload.cont:
9215 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9216 // CHECK14-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
9217 // CHECK14-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
9218 // CHECK14-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
9219 // CHECK14: omp_offload.failed2:
9220 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
9221 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT3]]
9222 // CHECK14: omp_offload.cont3:
9223 // CHECK14-NEXT: ret void
9224 //
9225 //
9226 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
9227 // CHECK14-SAME: () #[[ATTR1:[0-9]+]] {
9228 // CHECK14-NEXT: entry:
9229 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
9230 // CHECK14-NEXT: ret void
9231 //
9232 //
9233 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined.
9234 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9235 // CHECK14-NEXT: entry:
9236 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9237 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9238 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9239 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9240 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9241 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9242 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9243 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9244 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9245 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9246 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9247 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9248 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9249 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9250 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9251 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9252 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9253 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9254 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9255 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9256 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9257 // CHECK14: cond.true:
9258 // CHECK14-NEXT: br label [[COND_END:%.*]]
9259 // CHECK14: cond.false:
9260 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9261 // CHECK14-NEXT: br label [[COND_END]]
9262 // CHECK14: cond.end:
9263 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9264 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9265 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9266 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9267 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9268 // CHECK14: omp.inner.for.cond:
9269 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9270 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9271 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9272 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9273 // CHECK14: omp.inner.for.body:
9274 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9275 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9276 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9277 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9278 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
9279 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9280 // CHECK14: omp.inner.for.inc:
9281 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9282 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9283 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
9284 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9285 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9286 // CHECK14: omp.inner.for.end:
9287 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9288 // CHECK14: omp.loop.exit:
9289 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9290 // CHECK14-NEXT: ret void
9291 //
9292 //
9293 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..1
9294 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9295 // CHECK14-NEXT: entry:
9296 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9297 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9298 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9299 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9300 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9301 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9302 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9303 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9304 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9305 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9306 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9307 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9308 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9309 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9310 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9311 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9312 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9313 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9314 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9315 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9316 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9317 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9318 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9319 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9320 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9321 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9322 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9323 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9324 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9325 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9326 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9327 // CHECK14: cond.true:
9328 // CHECK14-NEXT: br label [[COND_END:%.*]]
9329 // CHECK14: cond.false:
9330 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9331 // CHECK14-NEXT: br label [[COND_END]]
9332 // CHECK14: cond.end:
9333 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9334 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9335 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9336 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9337 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9338 // CHECK14: omp.inner.for.cond:
9339 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9340 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9341 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9342 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9343 // CHECK14: omp.inner.for.body:
9344 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9345 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9346 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9347 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9348 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9349 // CHECK14: omp.body.continue:
9350 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9351 // CHECK14: omp.inner.for.inc:
9352 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9353 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9354 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9355 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9356 // CHECK14: omp.inner.for.end:
9357 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9358 // CHECK14: omp.loop.exit:
9359 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9360 // CHECK14-NEXT: ret void
9361 //
9362 //
9363 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
9364 // CHECK14-SAME: () #[[ATTR1]] {
9365 // CHECK14-NEXT: entry:
9366 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
9367 // CHECK14-NEXT: ret void
9368 //
9369 //
9370 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..2
9371 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9372 // CHECK14-NEXT: entry:
9373 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9374 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9375 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9376 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9377 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9378 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9379 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9380 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9381 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9382 // CHECK14-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9383 // CHECK14-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
9384 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9385 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9386 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9387 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9388 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9389 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9390 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9391 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9392 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9393 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9394 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9395 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9396 // CHECK14: cond.true:
9397 // CHECK14-NEXT: br label [[COND_END:%.*]]
9398 // CHECK14: cond.false:
9399 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9400 // CHECK14-NEXT: br label [[COND_END]]
9401 // CHECK14: cond.end:
9402 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9403 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9404 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9405 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9406 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9407 // CHECK14: omp.inner.for.cond:
9408 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9409 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9410 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9411 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9412 // CHECK14: omp.inner.for.body:
9413 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9414 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9415 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9416 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9417 // CHECK14-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
9418 // CHECK14-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9419 // CHECK14-NEXT: call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
9420 // CHECK14-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
9421 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9422 // CHECK14: omp.inner.for.inc:
9423 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9424 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9425 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9426 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9427 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9428 // CHECK14: omp.inner.for.end:
9429 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9430 // CHECK14: omp.loop.exit:
9431 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9432 // CHECK14-NEXT: ret void
9433 //
9434 //
9435 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..3
9436 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9437 // CHECK14-NEXT: entry:
9438 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9439 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9440 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9441 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9442 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9443 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9444 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9445 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9446 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9447 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9448 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9449 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9450 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9451 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9452 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9453 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9454 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9455 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9456 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9457 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9458 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9459 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9460 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9461 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9462 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9463 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9464 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9465 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9466 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9467 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9468 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9469 // CHECK14: cond.true:
9470 // CHECK14-NEXT: br label [[COND_END:%.*]]
9471 // CHECK14: cond.false:
9472 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9473 // CHECK14-NEXT: br label [[COND_END]]
9474 // CHECK14: cond.end:
9475 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9476 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9477 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9478 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9479 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9480 // CHECK14: omp.inner.for.cond:
9481 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9482 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9483 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9484 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9485 // CHECK14: omp.inner.for.body:
9486 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9487 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9488 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9489 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9490 // CHECK14-NEXT: call void @_Z9gtid_testv()
9491 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9492 // CHECK14: omp.body.continue:
9493 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9494 // CHECK14: omp.inner.for.inc:
9495 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9496 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9497 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9498 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9499 // CHECK14: omp.inner.for.end:
9500 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9501 // CHECK14: omp.loop.exit:
9502 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9503 // CHECK14-NEXT: ret void
9504 //
9505 //
9506 // CHECK14-LABEL: define {{[^@]+}}@main
9507 // CHECK14-SAME: () #[[ATTR3:[0-9]+]] {
9508 // CHECK14-NEXT: entry:
9509 // CHECK14-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
9510 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9511 // CHECK14-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
9512 // CHECK14-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
9513 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
9514 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
9515 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
9516 // CHECK14-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
9517 // CHECK14-NEXT: store i32 0, i32* [[RETVAL]], align 4
9518 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9519 // CHECK14-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
9520 // CHECK14-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
9521 // CHECK14-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9522 // CHECK14: omp_offload.failed:
9523 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
9524 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT]]
9525 // CHECK14: omp_offload.cont:
9526 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9527 // CHECK14-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
9528 // CHECK14-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
9529 // CHECK14-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
9530 // CHECK14: omp_offload.failed2:
9531 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
9532 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT3]]
9533 // CHECK14: omp_offload.cont3:
9534 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* @Arg, align 4
9535 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
9536 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
9537 // CHECK14-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
9538 // CHECK14-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9539 // CHECK14-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
9540 // CHECK14-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
9541 // CHECK14-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9542 // CHECK14-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
9543 // CHECK14-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
9544 // CHECK14-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9545 // CHECK14-NEXT: store i8* null, i8** [[TMP10]], align 8
9546 // CHECK14-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9547 // CHECK14-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9548 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9549 // CHECK14-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9550 // CHECK14-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9551 // CHECK14-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
9552 // CHECK14: omp_offload.failed5:
9553 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
9554 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT6]]
9555 // CHECK14: omp_offload.cont6:
9556 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* @Arg, align 4
9557 // CHECK14-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiEiT_(i32 [[TMP15]])
9558 // CHECK14-NEXT: ret i32 [[CALL]]
9559 //
9560 //
9561 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
9562 // CHECK14-SAME: () #[[ATTR1]] {
9563 // CHECK14-NEXT: entry:
9564 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
9565 // CHECK14-NEXT: ret void
9566 //
9567 //
9568 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..4
9569 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9570 // CHECK14-NEXT: entry:
9571 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9572 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9573 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9574 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9575 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9576 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9577 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9578 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9579 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9580 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9581 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9582 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9583 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9584 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9585 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9586 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9587 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9588 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9589 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9590 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9591 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9592 // CHECK14: cond.true:
9593 // CHECK14-NEXT: br label [[COND_END:%.*]]
9594 // CHECK14: cond.false:
9595 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9596 // CHECK14-NEXT: br label [[COND_END]]
9597 // CHECK14: cond.end:
9598 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9599 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9600 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9601 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9602 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9603 // CHECK14: omp.inner.for.cond:
9604 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9605 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9606 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9607 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9608 // CHECK14: omp.inner.for.body:
9609 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9610 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9611 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9612 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9613 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
9614 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9615 // CHECK14: omp.inner.for.inc:
9616 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9617 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9618 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
9619 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9620 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9621 // CHECK14: omp.inner.for.end:
9622 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9623 // CHECK14: omp.loop.exit:
9624 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9625 // CHECK14-NEXT: ret void
9626 //
9627 //
9628 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..5
9629 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9630 // CHECK14-NEXT: entry:
9631 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9632 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9633 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9634 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9635 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9636 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9637 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9638 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9639 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9640 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9641 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9642 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9643 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9644 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9645 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9646 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9647 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9648 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9649 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9650 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9651 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9652 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9653 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9654 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9655 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9656 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9657 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9658 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9659 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9660 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9661 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9662 // CHECK14: cond.true:
9663 // CHECK14-NEXT: br label [[COND_END:%.*]]
9664 // CHECK14: cond.false:
9665 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9666 // CHECK14-NEXT: br label [[COND_END]]
9667 // CHECK14: cond.end:
9668 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9669 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9670 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9671 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9672 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9673 // CHECK14: omp.inner.for.cond:
9674 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9675 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9676 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9677 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9678 // CHECK14: omp.inner.for.body:
9679 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9680 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9681 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9682 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9683 // CHECK14-NEXT: call void @_Z3fn4v()
9684 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9685 // CHECK14: omp.body.continue:
9686 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9687 // CHECK14: omp.inner.for.inc:
9688 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9689 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9690 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9691 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9692 // CHECK14: omp.inner.for.end:
9693 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9694 // CHECK14: omp.loop.exit:
9695 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9696 // CHECK14-NEXT: ret void
9697 //
9698 //
9699 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
9700 // CHECK14-SAME: () #[[ATTR1]] {
9701 // CHECK14-NEXT: entry:
9702 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
9703 // CHECK14-NEXT: ret void
9704 //
9705 //
9706 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..6
9707 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9708 // CHECK14-NEXT: entry:
9709 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9710 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9711 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9712 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9713 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9714 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9715 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9716 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9717 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9718 // CHECK14-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9719 // CHECK14-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
9720 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9721 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9722 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9723 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9724 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9725 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9726 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9727 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9728 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9729 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9730 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9731 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9732 // CHECK14: cond.true:
9733 // CHECK14-NEXT: br label [[COND_END:%.*]]
9734 // CHECK14: cond.false:
9735 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9736 // CHECK14-NEXT: br label [[COND_END]]
9737 // CHECK14: cond.end:
9738 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9739 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9740 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9741 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9742 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9743 // CHECK14: omp.inner.for.cond:
9744 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9745 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9746 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9747 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9748 // CHECK14: omp.inner.for.body:
9749 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9750 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9751 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9752 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9753 // CHECK14-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
9754 // CHECK14-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9755 // CHECK14-NEXT: call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
9756 // CHECK14-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
9757 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9758 // CHECK14: omp.inner.for.inc:
9759 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9760 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9761 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9762 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9763 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9764 // CHECK14: omp.inner.for.end:
9765 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9766 // CHECK14: omp.loop.exit:
9767 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9768 // CHECK14-NEXT: ret void
9769 //
9770 //
9771 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..7
9772 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9773 // CHECK14-NEXT: entry:
9774 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9775 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9776 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9777 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9778 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9779 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9780 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9781 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9782 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9783 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9784 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9785 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9786 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9787 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9788 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9789 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9790 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9791 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9792 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9793 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9794 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9795 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9796 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9797 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9798 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9799 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9800 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9801 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9802 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9803 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9804 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9805 // CHECK14: cond.true:
9806 // CHECK14-NEXT: br label [[COND_END:%.*]]
9807 // CHECK14: cond.false:
9808 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9809 // CHECK14-NEXT: br label [[COND_END]]
9810 // CHECK14: cond.end:
9811 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9812 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9813 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9814 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9815 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9816 // CHECK14: omp.inner.for.cond:
9817 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9818 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9819 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9820 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9821 // CHECK14: omp.inner.for.body:
9822 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9823 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9824 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9825 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9826 // CHECK14-NEXT: call void @_Z3fn5v()
9827 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9828 // CHECK14: omp.body.continue:
9829 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9830 // CHECK14: omp.inner.for.inc:
9831 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9832 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9833 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9834 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9835 // CHECK14: omp.inner.for.end:
9836 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9837 // CHECK14: omp.loop.exit:
9838 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9839 // CHECK14-NEXT: ret void
9840 //
9841 //
9842 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
9843 // CHECK14-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
9844 // CHECK14-NEXT: entry:
9845 // CHECK14-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
9846 // CHECK14-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
9847 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
9848 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
9849 // CHECK14-NEXT: ret void
9850 //
9851 //
9852 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..8
9853 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
9854 // CHECK14-NEXT: entry:
9855 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9856 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9857 // CHECK14-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
9858 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9859 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9860 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9861 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9862 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9863 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9864 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9865 // CHECK14-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9866 // CHECK14-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
9867 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9868 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9869 // CHECK14-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
9870 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
9871 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9872 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9873 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9874 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9875 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9876 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9877 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9878 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9879 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
9880 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9881 // CHECK14: cond.true:
9882 // CHECK14-NEXT: br label [[COND_END:%.*]]
9883 // CHECK14: cond.false:
9884 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9885 // CHECK14-NEXT: br label [[COND_END]]
9886 // CHECK14: cond.end:
9887 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9888 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9889 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9890 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9891 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9892 // CHECK14: omp.inner.for.cond:
9893 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9894 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9895 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9896 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9897 // CHECK14: omp.inner.for.body:
9898 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9899 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9900 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9901 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9902 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
9903 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
9904 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9905 // CHECK14: omp_if.then:
9906 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
9907 // CHECK14-NEXT: br label [[OMP_IF_END:%.*]]
9908 // CHECK14: omp_if.else:
9909 // CHECK14-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
9910 // CHECK14-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9911 // CHECK14-NEXT: call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
9912 // CHECK14-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
9913 // CHECK14-NEXT: br label [[OMP_IF_END]]
9914 // CHECK14: omp_if.end:
9915 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9916 // CHECK14: omp.inner.for.inc:
9917 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9918 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9919 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9920 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9921 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9922 // CHECK14: omp.inner.for.end:
9923 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9924 // CHECK14: omp.loop.exit:
9925 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9926 // CHECK14-NEXT: ret void
9927 //
9928 //
9929 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..9
9930 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9931 // CHECK14-NEXT: entry:
9932 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9933 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9934 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9935 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9936 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9937 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9938 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9939 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9940 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9941 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9942 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9943 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9944 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9945 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9946 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9947 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9948 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
9949 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9950 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9951 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9952 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9953 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9954 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9955 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9956 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9957 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9958 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9959 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9960 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9961 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9962 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9963 // CHECK14: cond.true:
9964 // CHECK14-NEXT: br label [[COND_END:%.*]]
9965 // CHECK14: cond.false:
9966 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9967 // CHECK14-NEXT: br label [[COND_END]]
9968 // CHECK14: cond.end:
9969 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9970 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9971 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9972 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9973 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9974 // CHECK14: omp.inner.for.cond:
9975 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9976 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9977 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9978 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9979 // CHECK14: omp.inner.for.body:
9980 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9981 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9982 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9983 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9984 // CHECK14-NEXT: call void @_Z3fn6v()
9985 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9986 // CHECK14: omp.body.continue:
9987 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9988 // CHECK14: omp.inner.for.inc:
9989 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9990 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9991 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9992 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
9993 // CHECK14: omp.inner.for.end:
9994 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9995 // CHECK14: omp.loop.exit:
9996 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9997 // CHECK14-NEXT: ret void
9998 //
9999 //
10000 // CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
10001 // CHECK14-SAME: (i32 [[ARG:%.*]]) #[[ATTR0]] comdat {
10002 // CHECK14-NEXT: entry:
10003 // CHECK14-NEXT: [[ARG_ADDR:%.*]] = alloca i32, align 4
10004 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10005 // CHECK14-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
10006 // CHECK14-NEXT: [[ARG_CASTED:%.*]] = alloca i64, align 8
10007 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
10008 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
10009 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
10010 // CHECK14-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
10011 // CHECK14-NEXT: store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
10012 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10013 // CHECK14-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
10014 // CHECK14-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
10015 // CHECK14-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10016 // CHECK14: omp_offload.failed:
10017 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
10018 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT]]
10019 // CHECK14: omp_offload.cont:
10020 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10021 // CHECK14-NEXT: [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
10022 // CHECK14-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
10023 // CHECK14-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
10024 // CHECK14: omp_offload.failed2:
10025 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
10026 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT3]]
10027 // CHECK14: omp_offload.cont3:
10028 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
10029 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
10030 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[CONV]], align 4
10031 // CHECK14-NEXT: [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
10032 // CHECK14-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10033 // CHECK14-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
10034 // CHECK14-NEXT: store i64 [[TMP5]], i64* [[TMP7]], align 8
10035 // CHECK14-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10036 // CHECK14-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
10037 // CHECK14-NEXT: store i64 [[TMP5]], i64* [[TMP9]], align 8
10038 // CHECK14-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10039 // CHECK14-NEXT: store i8* null, i8** [[TMP10]], align 8
10040 // CHECK14-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10041 // CHECK14-NEXT: [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10042 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10043 // CHECK14-NEXT: [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10044 // CHECK14-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
10045 // CHECK14-NEXT: br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
10046 // CHECK14: omp_offload.failed5:
10047 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
10048 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT6]]
10049 // CHECK14: omp_offload.cont6:
10050 // CHECK14-NEXT: ret i32 0
10051 //
10052 //
10053 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
10054 // CHECK14-SAME: () #[[ATTR1]] {
10055 // CHECK14-NEXT: entry:
10056 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
10057 // CHECK14-NEXT: ret void
10058 //
10059 //
10060 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..10
10061 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10062 // CHECK14-NEXT: entry:
10063 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10064 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10065 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10066 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10067 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10068 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10069 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10070 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10071 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10072 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10073 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10074 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10075 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10076 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10077 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10078 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10079 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10080 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10081 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10082 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10083 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10084 // CHECK14: cond.true:
10085 // CHECK14-NEXT: br label [[COND_END:%.*]]
10086 // CHECK14: cond.false:
10087 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10088 // CHECK14-NEXT: br label [[COND_END]]
10089 // CHECK14: cond.end:
10090 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10091 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10092 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10093 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10094 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10095 // CHECK14: omp.inner.for.cond:
10096 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10097 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10098 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10099 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10100 // CHECK14: omp.inner.for.body:
10101 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10102 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10103 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10104 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10105 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
10106 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10107 // CHECK14: omp.inner.for.inc:
10108 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10109 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10110 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
10111 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10112 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10113 // CHECK14: omp.inner.for.end:
10114 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10115 // CHECK14: omp.loop.exit:
10116 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10117 // CHECK14-NEXT: ret void
10118 //
10119 //
10120 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..11
10121 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10122 // CHECK14-NEXT: entry:
10123 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10124 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10125 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10126 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10127 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10128 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10129 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10130 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10131 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10132 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10133 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10134 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10135 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10136 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10137 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10138 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10139 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
10140 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10141 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10142 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10143 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10144 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10145 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10146 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10147 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10148 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10149 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10150 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10151 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10152 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10153 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10154 // CHECK14: cond.true:
10155 // CHECK14-NEXT: br label [[COND_END:%.*]]
10156 // CHECK14: cond.false:
10157 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10158 // CHECK14-NEXT: br label [[COND_END]]
10159 // CHECK14: cond.end:
10160 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10161 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10162 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10163 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10164 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10165 // CHECK14: omp.inner.for.cond:
10166 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10167 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10168 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10169 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10170 // CHECK14: omp.inner.for.body:
10171 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10172 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10173 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10174 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
10175 // CHECK14-NEXT: call void @_Z3fn1v()
10176 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10177 // CHECK14: omp.body.continue:
10178 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10179 // CHECK14: omp.inner.for.inc:
10180 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10181 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10182 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
10183 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10184 // CHECK14: omp.inner.for.end:
10185 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10186 // CHECK14: omp.loop.exit:
10187 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10188 // CHECK14-NEXT: ret void
10189 //
10190 //
10191 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
10192 // CHECK14-SAME: () #[[ATTR1]] {
10193 // CHECK14-NEXT: entry:
10194 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
10195 // CHECK14-NEXT: ret void
10196 //
10197 //
10198 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..12
10199 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10200 // CHECK14-NEXT: entry:
10201 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10202 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10203 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10204 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10205 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10206 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10207 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10208 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10209 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10210 // CHECK14-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10211 // CHECK14-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
10212 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10213 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10214 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10215 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10216 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10217 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10218 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10219 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10220 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10221 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10222 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10223 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10224 // CHECK14: cond.true:
10225 // CHECK14-NEXT: br label [[COND_END:%.*]]
10226 // CHECK14: cond.false:
10227 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10228 // CHECK14-NEXT: br label [[COND_END]]
10229 // CHECK14: cond.end:
10230 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10231 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10232 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10233 // CHECK14-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10234 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10235 // CHECK14: omp.inner.for.cond:
10236 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10237 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10238 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10239 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10240 // CHECK14: omp.inner.for.body:
10241 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10242 // CHECK14-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10243 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10244 // CHECK14-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10245 // CHECK14-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
10246 // CHECK14-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10247 // CHECK14-NEXT: call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
10248 // CHECK14-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
10249 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10250 // CHECK14: omp.inner.for.inc:
10251 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10252 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10253 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
10254 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10255 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10256 // CHECK14: omp.inner.for.end:
10257 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10258 // CHECK14: omp.loop.exit:
10259 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10260 // CHECK14-NEXT: ret void
10261 //
10262 //
10263 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..13
10264 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10265 // CHECK14-NEXT: entry:
10266 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10267 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10268 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10269 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10270 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10271 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10272 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10273 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10274 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10275 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10276 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10277 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10278 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10279 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10280 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10281 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10282 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
10283 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10284 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10285 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10286 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10287 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10288 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10289 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10290 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10291 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10292 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10293 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10294 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10295 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10296 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10297 // CHECK14: cond.true:
10298 // CHECK14-NEXT: br label [[COND_END:%.*]]
10299 // CHECK14: cond.false:
10300 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10301 // CHECK14-NEXT: br label [[COND_END]]
10302 // CHECK14: cond.end:
10303 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10304 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10305 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10306 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10307 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10308 // CHECK14: omp.inner.for.cond:
10309 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10310 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10311 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10312 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10313 // CHECK14: omp.inner.for.body:
10314 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10315 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10316 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10317 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
10318 // CHECK14-NEXT: call void @_Z3fn2v()
10319 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10320 // CHECK14: omp.body.continue:
10321 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10322 // CHECK14: omp.inner.for.inc:
10323 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10324 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10325 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
10326 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10327 // CHECK14: omp.inner.for.end:
10328 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10329 // CHECK14: omp.loop.exit:
10330 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10331 // CHECK14-NEXT: ret void
10332 //
10333 //
10334 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
10335 // CHECK14-SAME: (i64 [[ARG:%.*]]) #[[ATTR1]] {
10336 // CHECK14-NEXT: entry:
10337 // CHECK14-NEXT: [[ARG_ADDR:%.*]] = alloca i64, align 8
10338 // CHECK14-NEXT: store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
10339 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
10340 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
10341 // CHECK14-NEXT: ret void
10342 //
10343 //
10344 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..14
10345 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
10346 // CHECK14-NEXT: entry:
10347 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10348 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10349 // CHECK14-NEXT: [[ARG_ADDR:%.*]] = alloca i32*, align 8
10350 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10351 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10352 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10353 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10354 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10355 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10356 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10357 // CHECK14-NEXT: [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10358 // CHECK14-NEXT: store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
10359 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10360 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10361 // CHECK14-NEXT: store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
10362 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
10363 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10364 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10365 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10366 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10367 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10368 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10369 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10370 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10371 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
10372 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10373 // CHECK14: cond.true:
10374 // CHECK14-NEXT: br label [[COND_END:%.*]]
10375 // CHECK14: cond.false:
10376 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10377 // CHECK14-NEXT: br label [[COND_END]]
10378 // CHECK14: cond.end:
10379 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
10380 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10381 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10382 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
10383 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10384 // CHECK14: omp.inner.for.cond:
10385 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10386 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10387 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
10388 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10389 // CHECK14: omp.inner.for.body:
10390 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10391 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
10392 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10393 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
10394 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
10395 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
10396 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10397 // CHECK14: omp_if.then:
10398 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]])
10399 // CHECK14-NEXT: br label [[OMP_IF_END:%.*]]
10400 // CHECK14: omp_if.else:
10401 // CHECK14-NEXT: call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
10402 // CHECK14-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10403 // CHECK14-NEXT: call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
10404 // CHECK14-NEXT: call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
10405 // CHECK14-NEXT: br label [[OMP_IF_END]]
10406 // CHECK14: omp_if.end:
10407 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10408 // CHECK14: omp.inner.for.inc:
10409 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10410 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10411 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
10412 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10413 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10414 // CHECK14: omp.inner.for.end:
10415 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10416 // CHECK14: omp.loop.exit:
10417 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
10418 // CHECK14-NEXT: ret void
10419 //
10420 //
10421 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..15
10422 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10423 // CHECK14-NEXT: entry:
10424 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10425 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10426 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10427 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10428 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10429 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10430 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10431 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10432 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10433 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10434 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10435 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10436 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10437 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10438 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10439 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10440 // CHECK14-NEXT: store i32 99, i32* [[DOTOMP_UB]], align 4
10441 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10442 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10443 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10444 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10445 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10446 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10447 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10448 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10449 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10450 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10451 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10452 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10453 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10454 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10455 // CHECK14: cond.true:
10456 // CHECK14-NEXT: br label [[COND_END:%.*]]
10457 // CHECK14: cond.false:
10458 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10459 // CHECK14-NEXT: br label [[COND_END]]
10460 // CHECK14: cond.end:
10461 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10462 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10463 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10464 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10465 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10466 // CHECK14: omp.inner.for.cond:
10467 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10468 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10469 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10470 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10471 // CHECK14: omp.inner.for.body:
10472 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10473 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10474 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10475 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
10476 // CHECK14-NEXT: call void @_Z3fn3v()
10477 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10478 // CHECK14: omp.body.continue:
10479 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10480 // CHECK14: omp.inner.for.inc:
10481 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10482 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10483 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
10484 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10485 // CHECK14: omp.inner.for.end:
10486 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10487 // CHECK14: omp.loop.exit:
10488 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10489 // CHECK14-NEXT: ret void
10490 //
10491 //
10492 // CHECK14-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
10493 // CHECK14-SAME: () #[[ATTR5:[0-9]+]] {
10494 // CHECK14-NEXT: entry:
10495 // CHECK14-NEXT: call void @__tgt_register_requires(i64 1)
10496 // CHECK14-NEXT: ret void
10497 //
10498 //