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 // add -fopenmp-targets
3
4 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
5 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
6 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
7
8 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK4
11 // expected-no-diagnostics
12 #ifndef HEADER
13 #define HEADER
14
15 typedef __INTPTR_TYPE__ intptr_t;
16
17
18 void foo();
19
20 struct S {
21 intptr_t a, b, c;
SS22 S(intptr_t a) : a(a) {}
operator charS23 operator char() { return a; }
~SS24 ~S() {}
25 };
26
27 template <typename T>
tmain()28 T tmain() {
29 #pragma omp target
30 #pragma omp teams
31 #pragma omp distribute parallel for simd proc_bind(master)
32 for(int i = 0; i < 1000; i++) {}
33 return T();
34 }
35
main()36 int main() {
37 #pragma omp target
38 #pragma omp teams
39 #pragma omp distribute parallel for simd proc_bind(spread)
40 for(int i = 0; i < 1000; i++) {}
41 #pragma omp target
42 #pragma omp teams
43 #pragma omp distribute parallel for simd proc_bind(close)
44 for(int i = 0; i < 1000; i++) {}
45 return tmain<int>();
46 }
47
48
49
50
51
52
53
54
55 #endif
56 // CHECK1-LABEL: define {{[^@]+}}@main
57 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
58 // CHECK1-NEXT: entry:
59 // CHECK1-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
60 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
61 // CHECK1-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
62 // CHECK1-NEXT: store i32 0, i32* [[RETVAL]], align 4
63 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
64 // 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_l37.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
65 // CHECK1-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
66 // CHECK1-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
67 // CHECK1: omp_offload.failed:
68 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37() #[[ATTR2:[0-9]+]]
69 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
70 // CHECK1: omp_offload.cont:
71 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
72 // 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_l41.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
73 // CHECK1-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
74 // CHECK1-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
75 // CHECK1: omp_offload.failed2:
76 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41() #[[ATTR2]]
77 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT3]]
78 // CHECK1: omp_offload.cont3:
79 // CHECK1-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
80 // CHECK1-NEXT: ret i32 [[CALL]]
81 //
82 //
83 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
84 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
85 // CHECK1-NEXT: entry:
86 // 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*, ...)*))
87 // CHECK1-NEXT: ret void
88 //
89 //
90 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
91 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
92 // CHECK1-NEXT: entry:
93 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
94 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
95 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
98 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
99 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
100 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
101 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
102 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
103 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
104 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
105 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
106 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
107 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
108 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
109 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
110 // 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)
111 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
112 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
113 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
114 // CHECK1: cond.true:
115 // CHECK1-NEXT: br label [[COND_END:%.*]]
116 // CHECK1: cond.false:
117 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
118 // CHECK1-NEXT: br label [[COND_END]]
119 // CHECK1: cond.end:
120 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
121 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
122 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
123 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
124 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
125 // CHECK1: omp.inner.for.cond:
126 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
127 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
128 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
129 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
130 // CHECK1: omp.inner.for.body:
131 // CHECK1-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
132 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
133 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
134 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
135 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
136 // 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]])
137 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
138 // CHECK1: omp.inner.for.inc:
139 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
140 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
141 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
142 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
143 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
144 // CHECK1: omp.inner.for.end:
145 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
146 // CHECK1: omp.loop.exit:
147 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
148 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
149 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
150 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
151 // CHECK1: .omp.final.then:
152 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
153 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
154 // CHECK1: .omp.final.done:
155 // CHECK1-NEXT: ret void
156 //
157 //
158 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
159 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
160 // CHECK1-NEXT: entry:
161 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
162 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
163 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
164 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
165 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
167 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
168 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
169 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
170 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
171 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
172 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
173 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
174 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
175 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
176 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
177 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
178 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
179 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
180 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
181 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
182 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
183 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
184 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
185 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
186 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
187 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
188 // 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)
189 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
190 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
191 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
192 // CHECK1: cond.true:
193 // CHECK1-NEXT: br label [[COND_END:%.*]]
194 // CHECK1: cond.false:
195 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
196 // CHECK1-NEXT: br label [[COND_END]]
197 // CHECK1: cond.end:
198 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
199 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
200 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
201 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
202 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
203 // CHECK1: omp.inner.for.cond:
204 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
205 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
206 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
207 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
208 // CHECK1: omp.inner.for.body:
209 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
210 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
211 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
212 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
213 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
214 // CHECK1: omp.body.continue:
215 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
216 // CHECK1: omp.inner.for.inc:
217 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
218 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
219 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
220 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
221 // CHECK1: omp.inner.for.end:
222 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
223 // CHECK1: omp.loop.exit:
224 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
225 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
226 // CHECK1-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
227 // CHECK1-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
228 // CHECK1: .omp.final.then:
229 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
230 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
231 // CHECK1: .omp.final.done:
232 // CHECK1-NEXT: ret void
233 //
234 //
235 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
236 // CHECK1-SAME: () #[[ATTR1]] {
237 // CHECK1-NEXT: entry:
238 // 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*, ...)*))
239 // CHECK1-NEXT: ret void
240 //
241 //
242 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
243 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
244 // CHECK1-NEXT: entry:
245 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
246 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
247 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
248 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
249 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
250 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
251 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
252 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
253 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
254 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
255 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
256 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
257 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
258 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
259 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
260 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
261 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
262 // 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)
263 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
264 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
265 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
266 // CHECK1: cond.true:
267 // CHECK1-NEXT: br label [[COND_END:%.*]]
268 // CHECK1: cond.false:
269 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
270 // CHECK1-NEXT: br label [[COND_END]]
271 // CHECK1: cond.end:
272 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
273 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
274 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
275 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
276 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
277 // CHECK1: omp.inner.for.cond:
278 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
279 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
280 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
281 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
282 // CHECK1: omp.inner.for.body:
283 // CHECK1-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
284 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
285 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
286 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
287 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
288 // 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..3 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
289 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
290 // CHECK1: omp.inner.for.inc:
291 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
292 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
293 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
294 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
295 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
296 // CHECK1: omp.inner.for.end:
297 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
298 // CHECK1: omp.loop.exit:
299 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
300 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
301 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
302 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
303 // CHECK1: .omp.final.then:
304 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
305 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
306 // CHECK1: .omp.final.done:
307 // CHECK1-NEXT: ret void
308 //
309 //
310 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
311 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
312 // CHECK1-NEXT: entry:
313 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
314 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
315 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
316 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
317 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
318 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
320 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
321 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
322 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
323 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
324 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
325 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
326 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
327 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
328 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
329 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
330 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
331 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
332 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
333 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
334 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
335 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
336 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
337 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
338 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
339 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
340 // 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)
341 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
342 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
343 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
344 // CHECK1: cond.true:
345 // CHECK1-NEXT: br label [[COND_END:%.*]]
346 // CHECK1: cond.false:
347 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
348 // CHECK1-NEXT: br label [[COND_END]]
349 // CHECK1: cond.end:
350 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
351 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
352 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
353 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
354 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
355 // CHECK1: omp.inner.for.cond:
356 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
357 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
358 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
359 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
360 // CHECK1: omp.inner.for.body:
361 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
362 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
363 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
364 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
365 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
366 // CHECK1: omp.body.continue:
367 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
368 // CHECK1: omp.inner.for.inc:
369 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
370 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
371 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
372 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
373 // CHECK1: omp.inner.for.end:
374 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
375 // CHECK1: omp.loop.exit:
376 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
377 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
378 // CHECK1-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
379 // CHECK1-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
380 // CHECK1: .omp.final.then:
381 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
382 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
383 // CHECK1: .omp.final.done:
384 // CHECK1-NEXT: ret void
385 //
386 //
387 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
388 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
389 // CHECK1-NEXT: entry:
390 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
391 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
392 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
393 // CHECK1-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
394 // CHECK1-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
395 // CHECK1: omp_offload.failed:
396 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
397 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
398 // CHECK1: omp_offload.cont:
399 // CHECK1-NEXT: ret i32 0
400 //
401 //
402 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
403 // CHECK1-SAME: () #[[ATTR1]] {
404 // CHECK1-NEXT: entry:
405 // 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*, ...)*))
406 // CHECK1-NEXT: ret void
407 //
408 //
409 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
410 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
411 // CHECK1-NEXT: entry:
412 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
413 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
414 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
415 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
416 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
417 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
418 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
419 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
420 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
421 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
422 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
423 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
424 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
425 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
426 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
427 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
428 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
429 // 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)
430 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
431 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
432 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
433 // CHECK1: cond.true:
434 // CHECK1-NEXT: br label [[COND_END:%.*]]
435 // CHECK1: cond.false:
436 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
437 // CHECK1-NEXT: br label [[COND_END]]
438 // CHECK1: cond.end:
439 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
440 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
441 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
442 // CHECK1-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
443 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
444 // CHECK1: omp.inner.for.cond:
445 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
446 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
447 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
448 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
449 // CHECK1: omp.inner.for.body:
450 // CHECK1-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
451 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
452 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
453 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
454 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
455 // 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]])
456 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
457 // CHECK1: omp.inner.for.inc:
458 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
459 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
460 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
461 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
462 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
463 // CHECK1: omp.inner.for.end:
464 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
465 // CHECK1: omp.loop.exit:
466 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
467 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
468 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
469 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
470 // CHECK1: .omp.final.then:
471 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
472 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
473 // CHECK1: .omp.final.done:
474 // CHECK1-NEXT: ret void
475 //
476 //
477 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
478 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
479 // CHECK1-NEXT: entry:
480 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
481 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
482 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
483 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
484 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
485 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
486 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
487 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
488 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
489 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
490 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
491 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
492 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
493 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
494 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
495 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
496 // CHECK1-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
497 // CHECK1-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
498 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
499 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
500 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
501 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
502 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
503 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
504 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
505 // CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
506 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
507 // 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)
508 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
509 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
510 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
511 // CHECK1: cond.true:
512 // CHECK1-NEXT: br label [[COND_END:%.*]]
513 // CHECK1: cond.false:
514 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
515 // CHECK1-NEXT: br label [[COND_END]]
516 // CHECK1: cond.end:
517 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
518 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
519 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
520 // CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
521 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
522 // CHECK1: omp.inner.for.cond:
523 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
524 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
525 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
526 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
527 // CHECK1: omp.inner.for.body:
528 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
529 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
530 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
531 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
532 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
533 // CHECK1: omp.body.continue:
534 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
535 // CHECK1: omp.inner.for.inc:
536 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
537 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
538 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
539 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
540 // CHECK1: omp.inner.for.end:
541 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
542 // CHECK1: omp.loop.exit:
543 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
544 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
545 // CHECK1-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
546 // CHECK1-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
547 // CHECK1: .omp.final.then:
548 // CHECK1-NEXT: store i32 1000, i32* [[I]], align 4
549 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
550 // CHECK1: .omp.final.done:
551 // CHECK1-NEXT: ret void
552 //
553 //
554 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
555 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
556 // CHECK1-NEXT: entry:
557 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
558 // CHECK1-NEXT: ret void
559 //
560 //
561 // CHECK2-LABEL: define {{[^@]+}}@main
562 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
563 // CHECK2-NEXT: entry:
564 // CHECK2-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
565 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
566 // CHECK2-NEXT: [[_TMP1:%.*]] = alloca i32, align 4
567 // CHECK2-NEXT: store i32 0, i32* [[RETVAL]], align 4
568 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
569 // 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_l37.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
570 // CHECK2-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
571 // CHECK2-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
572 // CHECK2: omp_offload.failed:
573 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37() #[[ATTR2:[0-9]+]]
574 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
575 // CHECK2: omp_offload.cont:
576 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
577 // 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_l41.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
578 // CHECK2-NEXT: [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
579 // CHECK2-NEXT: br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
580 // CHECK2: omp_offload.failed2:
581 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41() #[[ATTR2]]
582 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT3]]
583 // CHECK2: omp_offload.cont3:
584 // CHECK2-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
585 // CHECK2-NEXT: ret i32 [[CALL]]
586 //
587 //
588 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l37
589 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
590 // CHECK2-NEXT: entry:
591 // 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*, ...)*))
592 // CHECK2-NEXT: ret void
593 //
594 //
595 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
596 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
597 // CHECK2-NEXT: entry:
598 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
599 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
600 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
601 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
602 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
603 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
604 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
605 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
606 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
607 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
608 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
609 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
610 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
611 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
612 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
613 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
614 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
615 // 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)
616 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
617 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
618 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
619 // CHECK2: cond.true:
620 // CHECK2-NEXT: br label [[COND_END:%.*]]
621 // CHECK2: cond.false:
622 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
623 // CHECK2-NEXT: br label [[COND_END]]
624 // CHECK2: cond.end:
625 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
626 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
627 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
628 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
629 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
630 // CHECK2: omp.inner.for.cond:
631 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
632 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
633 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
634 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
635 // CHECK2: omp.inner.for.body:
636 // CHECK2-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
637 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
638 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
639 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
640 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
641 // 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]])
642 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
643 // CHECK2: omp.inner.for.inc:
644 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
645 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
646 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
647 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
648 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
649 // CHECK2: omp.inner.for.end:
650 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
651 // CHECK2: omp.loop.exit:
652 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
653 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
654 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
655 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
656 // CHECK2: .omp.final.then:
657 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
658 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
659 // CHECK2: .omp.final.done:
660 // CHECK2-NEXT: ret void
661 //
662 //
663 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
664 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
665 // CHECK2-NEXT: entry:
666 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
667 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
668 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
669 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
670 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
671 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
672 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
673 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
674 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
675 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
676 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
677 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
678 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
679 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
680 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
681 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
682 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
683 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
684 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
685 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
686 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
687 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
688 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
689 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
690 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
691 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
692 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
693 // 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)
694 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
695 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
696 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
697 // CHECK2: cond.true:
698 // CHECK2-NEXT: br label [[COND_END:%.*]]
699 // CHECK2: cond.false:
700 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
701 // CHECK2-NEXT: br label [[COND_END]]
702 // CHECK2: cond.end:
703 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
704 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
705 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
706 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
707 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
708 // CHECK2: omp.inner.for.cond:
709 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
710 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
711 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
712 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
713 // CHECK2: omp.inner.for.body:
714 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
715 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
716 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
717 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
718 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
719 // CHECK2: omp.body.continue:
720 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
721 // CHECK2: omp.inner.for.inc:
722 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
723 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
724 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
725 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
726 // CHECK2: omp.inner.for.end:
727 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
728 // CHECK2: omp.loop.exit:
729 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
730 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
731 // CHECK2-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
732 // CHECK2-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
733 // CHECK2: .omp.final.then:
734 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
735 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
736 // CHECK2: .omp.final.done:
737 // CHECK2-NEXT: ret void
738 //
739 //
740 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l41
741 // CHECK2-SAME: () #[[ATTR1]] {
742 // CHECK2-NEXT: entry:
743 // 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*, ...)*))
744 // CHECK2-NEXT: ret void
745 //
746 //
747 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
748 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
749 // CHECK2-NEXT: entry:
750 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
751 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
752 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
753 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
754 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
755 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
756 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
757 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
758 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
759 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
760 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
761 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
762 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
763 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
764 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
765 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
766 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
767 // 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)
768 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
769 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
770 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
771 // CHECK2: cond.true:
772 // CHECK2-NEXT: br label [[COND_END:%.*]]
773 // CHECK2: cond.false:
774 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
775 // CHECK2-NEXT: br label [[COND_END]]
776 // CHECK2: cond.end:
777 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
778 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
779 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
780 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
781 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
782 // CHECK2: omp.inner.for.cond:
783 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
784 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
785 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
786 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
787 // CHECK2: omp.inner.for.body:
788 // CHECK2-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
789 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
790 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
791 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
792 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
793 // 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..3 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
794 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
795 // CHECK2: omp.inner.for.inc:
796 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
797 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
798 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
799 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
800 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
801 // CHECK2: omp.inner.for.end:
802 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
803 // CHECK2: omp.loop.exit:
804 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
805 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
806 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
807 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
808 // CHECK2: .omp.final.then:
809 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
810 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
811 // CHECK2: .omp.final.done:
812 // CHECK2-NEXT: ret void
813 //
814 //
815 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
816 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
817 // CHECK2-NEXT: entry:
818 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
819 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
820 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
821 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
822 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
823 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
824 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
825 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
826 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
827 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
828 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
829 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
830 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
831 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
832 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
833 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
834 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
835 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
836 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
837 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
838 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
839 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
840 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
841 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
842 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
843 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
844 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
845 // 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)
846 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
847 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
848 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
849 // CHECK2: cond.true:
850 // CHECK2-NEXT: br label [[COND_END:%.*]]
851 // CHECK2: cond.false:
852 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
853 // CHECK2-NEXT: br label [[COND_END]]
854 // CHECK2: cond.end:
855 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
856 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
857 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
858 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
859 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
860 // CHECK2: omp.inner.for.cond:
861 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
862 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
863 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
864 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
865 // CHECK2: omp.inner.for.body:
866 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
867 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
868 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
869 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
870 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
871 // CHECK2: omp.body.continue:
872 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
873 // CHECK2: omp.inner.for.inc:
874 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
875 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
876 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
877 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
878 // CHECK2: omp.inner.for.end:
879 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
880 // CHECK2: omp.loop.exit:
881 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
882 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
883 // CHECK2-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
884 // CHECK2-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
885 // CHECK2: .omp.final.then:
886 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
887 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
888 // CHECK2: .omp.final.done:
889 // CHECK2-NEXT: ret void
890 //
891 //
892 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
893 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat {
894 // CHECK2-NEXT: entry:
895 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
896 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
897 // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
898 // CHECK2-NEXT: [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
899 // CHECK2-NEXT: br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
900 // CHECK2: omp_offload.failed:
901 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
902 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
903 // CHECK2: omp_offload.cont:
904 // CHECK2-NEXT: ret i32 0
905 //
906 //
907 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
908 // CHECK2-SAME: () #[[ATTR1]] {
909 // CHECK2-NEXT: entry:
910 // 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*, ...)*))
911 // CHECK2-NEXT: ret void
912 //
913 //
914 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
915 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
916 // CHECK2-NEXT: entry:
917 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
918 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
919 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
920 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
921 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
922 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
923 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
924 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
925 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
926 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
927 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
928 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
929 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
930 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
931 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
932 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
933 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
934 // 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)
935 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
936 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
937 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
938 // CHECK2: cond.true:
939 // CHECK2-NEXT: br label [[COND_END:%.*]]
940 // CHECK2: cond.false:
941 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
942 // CHECK2-NEXT: br label [[COND_END]]
943 // CHECK2: cond.end:
944 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
945 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
946 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
947 // CHECK2-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
948 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
949 // CHECK2: omp.inner.for.cond:
950 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
951 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
952 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
953 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
954 // CHECK2: omp.inner.for.body:
955 // CHECK2-NEXT: call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
956 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
957 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
958 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
959 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
960 // 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]])
961 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
962 // CHECK2: omp.inner.for.inc:
963 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
964 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
965 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
966 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
967 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
968 // CHECK2: omp.inner.for.end:
969 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
970 // CHECK2: omp.loop.exit:
971 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
972 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
973 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
974 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
975 // CHECK2: .omp.final.then:
976 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
977 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
978 // CHECK2: .omp.final.done:
979 // CHECK2-NEXT: ret void
980 //
981 //
982 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
983 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
984 // CHECK2-NEXT: entry:
985 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
986 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
987 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
988 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
989 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
990 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
991 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
992 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
993 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
994 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
995 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
996 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
997 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
998 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
999 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1000 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1001 // CHECK2-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
1002 // CHECK2-NEXT: [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1003 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1004 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1005 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1006 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1007 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1008 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1009 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1010 // CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1011 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1012 // 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)
1013 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1014 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
1015 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1016 // CHECK2: cond.true:
1017 // CHECK2-NEXT: br label [[COND_END:%.*]]
1018 // CHECK2: cond.false:
1019 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1020 // CHECK2-NEXT: br label [[COND_END]]
1021 // CHECK2: cond.end:
1022 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1023 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1024 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1025 // CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1026 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1027 // CHECK2: omp.inner.for.cond:
1028 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1029 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1030 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1031 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1032 // CHECK2: omp.inner.for.body:
1033 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1034 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1035 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1036 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1037 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1038 // CHECK2: omp.body.continue:
1039 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1040 // CHECK2: omp.inner.for.inc:
1041 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1042 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1043 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1044 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1045 // CHECK2: omp.inner.for.end:
1046 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1047 // CHECK2: omp.loop.exit:
1048 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1049 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1050 // CHECK2-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1051 // CHECK2-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1052 // CHECK2: .omp.final.then:
1053 // CHECK2-NEXT: store i32 1000, i32* [[I]], align 4
1054 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1055 // CHECK2: .omp.final.done:
1056 // CHECK2-NEXT: ret void
1057 //
1058 //
1059 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1060 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
1061 // CHECK2-NEXT: entry:
1062 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1)
1063 // CHECK2-NEXT: ret void
1064 //
1065 //
1066 // CHECK3-LABEL: define {{[^@]+}}@main
1067 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1068 // CHECK3-NEXT: entry:
1069 // CHECK3-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1072 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1073 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1074 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1075 // CHECK3-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
1076 // CHECK3-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1077 // CHECK3-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1078 // CHECK3-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1079 // CHECK3-NEXT: [[I6:%.*]] = alloca i32, align 4
1080 // CHECK3-NEXT: store i32 0, i32* [[RETVAL]], align 4
1081 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1082 // CHECK3-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
1083 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1084 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1085 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1086 // CHECK3: omp.inner.for.cond:
1087 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1088 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1089 // CHECK3-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1090 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1091 // CHECK3: omp.inner.for.body:
1092 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1093 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1094 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1095 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1096 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1097 // CHECK3: omp.body.continue:
1098 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1099 // CHECK3: omp.inner.for.inc:
1100 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1101 // CHECK3-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1102 // CHECK3-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1103 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1104 // CHECK3: omp.inner.for.end:
1105 // CHECK3-NEXT: store i32 1000, i32* [[I]], align 4
1106 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
1107 // CHECK3-NEXT: store i32 999, i32* [[DOTOMP_UB4]], align 4
1108 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1109 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1110 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
1111 // CHECK3: omp.inner.for.cond7:
1112 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1113 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1114 // CHECK3-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1115 // CHECK3-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1116 // CHECK3: omp.inner.for.body9:
1117 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1118 // CHECK3-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1119 // CHECK3-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1120 // CHECK3-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1121 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE12:%.*]]
1122 // CHECK3: omp.body.continue12:
1123 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC13:%.*]]
1124 // CHECK3: omp.inner.for.inc13:
1125 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1126 // CHECK3-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1127 // CHECK3-NEXT: store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1128 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1129 // CHECK3: omp.inner.for.end15:
1130 // CHECK3-NEXT: store i32 1000, i32* [[I6]], align 4
1131 // CHECK3-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1132 // CHECK3-NEXT: ret i32 [[CALL]]
1133 //
1134 //
1135 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1136 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] comdat {
1137 // CHECK3-NEXT: entry:
1138 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1139 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1140 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1141 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1142 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1143 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1144 // CHECK3-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
1145 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1146 // CHECK3-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1147 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1148 // CHECK3: omp.inner.for.cond:
1149 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1150 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1151 // CHECK3-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1152 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1153 // CHECK3: omp.inner.for.body:
1154 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1155 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1156 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1157 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1158 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1159 // CHECK3: omp.body.continue:
1160 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1161 // CHECK3: omp.inner.for.inc:
1162 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1163 // CHECK3-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1164 // CHECK3-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1165 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1166 // CHECK3: omp.inner.for.end:
1167 // CHECK3-NEXT: store i32 1000, i32* [[I]], align 4
1168 // CHECK3-NEXT: ret i32 0
1169 //
1170 //
1171 // CHECK4-LABEL: define {{[^@]+}}@main
1172 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
1173 // CHECK4-NEXT: entry:
1174 // CHECK4-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1175 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1176 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1177 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1178 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1179 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
1180 // CHECK4-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
1181 // CHECK4-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1182 // CHECK4-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1183 // CHECK4-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1184 // CHECK4-NEXT: [[I6:%.*]] = alloca i32, align 4
1185 // CHECK4-NEXT: store i32 0, i32* [[RETVAL]], align 4
1186 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1187 // CHECK4-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
1188 // CHECK4-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1189 // CHECK4-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1190 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1191 // CHECK4: omp.inner.for.cond:
1192 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1193 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1194 // CHECK4-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1195 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1196 // CHECK4: omp.inner.for.body:
1197 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1198 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1199 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1200 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1201 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1202 // CHECK4: omp.body.continue:
1203 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1204 // CHECK4: omp.inner.for.inc:
1205 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1206 // CHECK4-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1207 // CHECK4-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1208 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1209 // CHECK4: omp.inner.for.end:
1210 // CHECK4-NEXT: store i32 1000, i32* [[I]], align 4
1211 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
1212 // CHECK4-NEXT: store i32 999, i32* [[DOTOMP_UB4]], align 4
1213 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1214 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1215 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
1216 // CHECK4: omp.inner.for.cond7:
1217 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1218 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1219 // CHECK4-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1220 // CHECK4-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1221 // CHECK4: omp.inner.for.body9:
1222 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1223 // CHECK4-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1224 // CHECK4-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1225 // CHECK4-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1226 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE12:%.*]]
1227 // CHECK4: omp.body.continue12:
1228 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC13:%.*]]
1229 // CHECK4: omp.inner.for.inc13:
1230 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1231 // CHECK4-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1232 // CHECK4-NEXT: store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1233 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1234 // CHECK4: omp.inner.for.end15:
1235 // CHECK4-NEXT: store i32 1000, i32* [[I6]], align 4
1236 // CHECK4-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1237 // CHECK4-NEXT: ret i32 [[CALL]]
1238 //
1239 //
1240 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1241 // CHECK4-SAME: () #[[ATTR1:[0-9]+]] comdat {
1242 // CHECK4-NEXT: entry:
1243 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1244 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1245 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1246 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1247 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
1248 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1249 // CHECK4-NEXT: store i32 999, i32* [[DOTOMP_UB]], align 4
1250 // CHECK4-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1251 // CHECK4-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1252 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1253 // CHECK4: omp.inner.for.cond:
1254 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1255 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1256 // CHECK4-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1257 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1258 // CHECK4: omp.inner.for.body:
1259 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1260 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1261 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1262 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1263 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1264 // CHECK4: omp.body.continue:
1265 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1266 // CHECK4: omp.inner.for.inc:
1267 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1268 // CHECK4-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1269 // CHECK4-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1270 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1271 // CHECK4: omp.inner.for.end:
1272 // CHECK4-NEXT: store i32 1000, i32* [[I]], align 4
1273 // CHECK4-NEXT: ret i32 0
1274 //
1275