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 distribute parallel for simd proc_bind(master)
31   for(int i = 0; i < 1000; i++) {}
32   return T();
33 }
34 
main()35 int main() {
36 #pragma omp target
37 #pragma omp teams distribute parallel for simd proc_bind(spread)
38   for(int i = 0; i < 1000; i++) {}
39 #pragma omp target
40 #pragma omp teams distribute parallel for simd proc_bind(close)
41   for(int i = 0; i < 1000; i++) {}
42   return tmain<int>();
43 }
44 
45 
46 
47 
48 
49 
50 
51 
52 
53 
54 #endif
55 // CHECK1-LABEL: define {{[^@]+}}@main
56 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
57 // CHECK1-NEXT:  entry:
58 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
59 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
60 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
61 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
62 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
63 // 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_l36.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
64 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
65 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
66 // CHECK1:       omp_offload.failed:
67 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
68 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
69 // CHECK1:       omp_offload.cont:
70 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
71 // 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_l39.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
72 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
73 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
74 // CHECK1:       omp_offload.failed2:
75 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39() #[[ATTR2]]
76 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
77 // CHECK1:       omp_offload.cont3:
78 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
79 // CHECK1-NEXT:    ret i32 [[CALL]]
80 //
81 //
82 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
83 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
84 // CHECK1-NEXT:  entry:
85 // 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*, ...)*))
86 // CHECK1-NEXT:    ret void
87 //
88 //
89 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
90 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
91 // CHECK1-NEXT:  entry:
92 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
93 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
94 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
95 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
98 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
99 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
100 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
101 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
102 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
103 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
104 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
105 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
106 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
107 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
108 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
109 // 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)
110 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
111 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
112 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
113 // CHECK1:       cond.true:
114 // CHECK1-NEXT:    br label [[COND_END:%.*]]
115 // CHECK1:       cond.false:
116 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
117 // CHECK1-NEXT:    br label [[COND_END]]
118 // CHECK1:       cond.end:
119 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
120 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
121 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
122 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
123 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
124 // CHECK1:       omp.inner.for.cond:
125 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
126 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
127 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
128 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
129 // CHECK1:       omp.inner.for.body:
130 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
131 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
132 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
133 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
134 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
135 // 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]])
136 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
137 // CHECK1:       omp.inner.for.inc:
138 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
139 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
140 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
141 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
142 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
143 // CHECK1:       omp.inner.for.end:
144 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
145 // CHECK1:       omp.loop.exit:
146 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
147 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
148 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
149 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
150 // CHECK1:       .omp.final.then:
151 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
152 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
153 // CHECK1:       .omp.final.done:
154 // CHECK1-NEXT:    ret void
155 //
156 //
157 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
158 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
159 // CHECK1-NEXT:  entry:
160 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
161 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
162 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
163 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
164 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
167 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
168 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
169 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
170 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
171 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
172 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
173 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
174 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
175 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
176 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
177 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
178 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
179 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
180 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
181 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
182 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
183 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
184 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
185 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
186 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
187 // 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)
188 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
189 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
190 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
191 // CHECK1:       cond.true:
192 // CHECK1-NEXT:    br label [[COND_END:%.*]]
193 // CHECK1:       cond.false:
194 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
195 // CHECK1-NEXT:    br label [[COND_END]]
196 // CHECK1:       cond.end:
197 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
198 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
199 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
200 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
201 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
202 // CHECK1:       omp.inner.for.cond:
203 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
204 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
205 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
206 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
207 // CHECK1:       omp.inner.for.body:
208 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
209 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
210 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
211 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
212 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
213 // CHECK1:       omp.body.continue:
214 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
215 // CHECK1:       omp.inner.for.inc:
216 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
217 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
218 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
219 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
220 // CHECK1:       omp.inner.for.end:
221 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
222 // CHECK1:       omp.loop.exit:
223 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
224 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
225 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
226 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
227 // CHECK1:       .omp.final.then:
228 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
229 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
230 // CHECK1:       .omp.final.done:
231 // CHECK1-NEXT:    ret void
232 //
233 //
234 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39
235 // CHECK1-SAME: () #[[ATTR1]] {
236 // CHECK1-NEXT:  entry:
237 // 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*, ...)*))
238 // CHECK1-NEXT:    ret void
239 //
240 //
241 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
242 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
243 // CHECK1-NEXT:  entry:
244 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
245 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
246 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
247 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
248 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
249 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
250 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
251 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
252 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
253 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
254 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
255 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
256 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
257 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
258 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
259 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
260 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
261 // 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)
262 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
263 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
264 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
265 // CHECK1:       cond.true:
266 // CHECK1-NEXT:    br label [[COND_END:%.*]]
267 // CHECK1:       cond.false:
268 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
269 // CHECK1-NEXT:    br label [[COND_END]]
270 // CHECK1:       cond.end:
271 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
272 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
273 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
274 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
275 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
276 // CHECK1:       omp.inner.for.cond:
277 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
278 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
279 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
280 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
281 // CHECK1:       omp.inner.for.body:
282 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
283 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
284 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
285 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
286 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
287 // 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]])
288 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
289 // CHECK1:       omp.inner.for.inc:
290 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
291 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
292 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
293 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
294 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
295 // CHECK1:       omp.inner.for.end:
296 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
297 // CHECK1:       omp.loop.exit:
298 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
299 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
300 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
301 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
302 // CHECK1:       .omp.final.then:
303 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
304 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
305 // CHECK1:       .omp.final.done:
306 // CHECK1-NEXT:    ret void
307 //
308 //
309 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
310 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
311 // CHECK1-NEXT:  entry:
312 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
313 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
314 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
315 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
316 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
317 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
318 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
320 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
321 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
322 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
323 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
324 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
325 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
326 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
327 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
328 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
329 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
330 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
331 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
332 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
333 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
334 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
335 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
336 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
337 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
338 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
339 // 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)
340 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
341 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
342 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
343 // CHECK1:       cond.true:
344 // CHECK1-NEXT:    br label [[COND_END:%.*]]
345 // CHECK1:       cond.false:
346 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
347 // CHECK1-NEXT:    br label [[COND_END]]
348 // CHECK1:       cond.end:
349 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
350 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
351 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
352 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
353 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
354 // CHECK1:       omp.inner.for.cond:
355 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
356 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
357 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
358 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
359 // CHECK1:       omp.inner.for.body:
360 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
361 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
362 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
363 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
364 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
365 // CHECK1:       omp.body.continue:
366 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
367 // CHECK1:       omp.inner.for.inc:
368 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
369 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
370 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
371 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
372 // CHECK1:       omp.inner.for.end:
373 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
374 // CHECK1:       omp.loop.exit:
375 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
376 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
377 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
378 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
379 // CHECK1:       .omp.final.then:
380 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
381 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
382 // CHECK1:       .omp.final.done:
383 // CHECK1-NEXT:    ret void
384 //
385 //
386 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
387 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
388 // CHECK1-NEXT:  entry:
389 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
390 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
391 // 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 0)
392 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
393 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
394 // CHECK1:       omp_offload.failed:
395 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
396 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
397 // CHECK1:       omp_offload.cont:
398 // CHECK1-NEXT:    ret i32 0
399 //
400 //
401 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
402 // CHECK1-SAME: () #[[ATTR1]] {
403 // CHECK1-NEXT:  entry:
404 // 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*, ...)*))
405 // CHECK1-NEXT:    ret void
406 //
407 //
408 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
409 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
410 // CHECK1-NEXT:  entry:
411 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
412 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
413 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
414 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
415 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
416 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
417 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
418 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
419 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
420 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
421 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
422 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
423 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
424 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
425 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
426 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
427 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
428 // 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)
429 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
430 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
431 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
432 // CHECK1:       cond.true:
433 // CHECK1-NEXT:    br label [[COND_END:%.*]]
434 // CHECK1:       cond.false:
435 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
436 // CHECK1-NEXT:    br label [[COND_END]]
437 // CHECK1:       cond.end:
438 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
439 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
440 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
441 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
442 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
443 // CHECK1:       omp.inner.for.cond:
444 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
445 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
446 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
447 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
448 // CHECK1:       omp.inner.for.body:
449 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
450 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
451 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
452 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
453 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
454 // 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]])
455 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
456 // CHECK1:       omp.inner.for.inc:
457 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
458 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
459 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
460 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
461 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
462 // CHECK1:       omp.inner.for.end:
463 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
464 // CHECK1:       omp.loop.exit:
465 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
466 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
467 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
468 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
469 // CHECK1:       .omp.final.then:
470 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
471 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
472 // CHECK1:       .omp.final.done:
473 // CHECK1-NEXT:    ret void
474 //
475 //
476 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
477 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
478 // CHECK1-NEXT:  entry:
479 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
480 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
481 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
482 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
483 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
484 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
485 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
486 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
487 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
488 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
489 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
490 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
491 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
492 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
493 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
494 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
495 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
496 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
497 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
498 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
499 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
500 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
501 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
502 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
503 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
504 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
505 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
506 // 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)
507 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
508 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
509 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
510 // CHECK1:       cond.true:
511 // CHECK1-NEXT:    br label [[COND_END:%.*]]
512 // CHECK1:       cond.false:
513 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
514 // CHECK1-NEXT:    br label [[COND_END]]
515 // CHECK1:       cond.end:
516 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
517 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
518 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
519 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
520 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
521 // CHECK1:       omp.inner.for.cond:
522 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
523 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
524 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
525 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
526 // CHECK1:       omp.inner.for.body:
527 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
528 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
529 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
530 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
531 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
532 // CHECK1:       omp.body.continue:
533 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
534 // CHECK1:       omp.inner.for.inc:
535 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
536 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
537 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
538 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
539 // CHECK1:       omp.inner.for.end:
540 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
541 // CHECK1:       omp.loop.exit:
542 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
543 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
544 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
545 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
546 // CHECK1:       .omp.final.then:
547 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
548 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
549 // CHECK1:       .omp.final.done:
550 // CHECK1-NEXT:    ret void
551 //
552 //
553 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
554 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
555 // CHECK1-NEXT:  entry:
556 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
557 // CHECK1-NEXT:    ret void
558 //
559 //
560 // CHECK2-LABEL: define {{[^@]+}}@main
561 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
562 // CHECK2-NEXT:  entry:
563 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
564 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
565 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
566 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
567 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
568 // 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_l36.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
569 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
570 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
571 // CHECK2:       omp_offload.failed:
572 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
573 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
574 // CHECK2:       omp_offload.cont:
575 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
576 // 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_l39.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
577 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
578 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
579 // CHECK2:       omp_offload.failed2:
580 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39() #[[ATTR2]]
581 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
582 // CHECK2:       omp_offload.cont3:
583 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
584 // CHECK2-NEXT:    ret i32 [[CALL]]
585 //
586 //
587 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
588 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
589 // CHECK2-NEXT:  entry:
590 // 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*, ...)*))
591 // CHECK2-NEXT:    ret void
592 //
593 //
594 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
595 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
596 // CHECK2-NEXT:  entry:
597 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
598 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
599 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
600 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
601 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
602 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
603 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
604 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
605 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
606 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
607 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
608 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
609 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
610 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
611 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
612 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
613 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
614 // 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)
615 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
616 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
617 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
618 // CHECK2:       cond.true:
619 // CHECK2-NEXT:    br label [[COND_END:%.*]]
620 // CHECK2:       cond.false:
621 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
622 // CHECK2-NEXT:    br label [[COND_END]]
623 // CHECK2:       cond.end:
624 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
625 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
626 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
627 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
628 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
629 // CHECK2:       omp.inner.for.cond:
630 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
631 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
632 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
633 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
634 // CHECK2:       omp.inner.for.body:
635 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
636 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
637 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
638 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
639 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
640 // 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]])
641 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
642 // CHECK2:       omp.inner.for.inc:
643 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
644 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
645 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
646 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
647 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
648 // CHECK2:       omp.inner.for.end:
649 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
650 // CHECK2:       omp.loop.exit:
651 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
652 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
653 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
654 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
655 // CHECK2:       .omp.final.then:
656 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
657 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
658 // CHECK2:       .omp.final.done:
659 // CHECK2-NEXT:    ret void
660 //
661 //
662 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
663 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
664 // CHECK2-NEXT:  entry:
665 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
666 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
667 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
668 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
669 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
670 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
671 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
672 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
673 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
674 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
675 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
676 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
677 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
678 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
679 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
680 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
681 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
682 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
683 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
684 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
685 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
686 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
687 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
688 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
689 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
690 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
691 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
692 // 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)
693 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
694 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
695 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
696 // CHECK2:       cond.true:
697 // CHECK2-NEXT:    br label [[COND_END:%.*]]
698 // CHECK2:       cond.false:
699 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
700 // CHECK2-NEXT:    br label [[COND_END]]
701 // CHECK2:       cond.end:
702 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
703 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
704 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
705 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
706 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
707 // CHECK2:       omp.inner.for.cond:
708 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
709 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
710 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
711 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
712 // CHECK2:       omp.inner.for.body:
713 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
714 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
715 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
716 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
717 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
718 // CHECK2:       omp.body.continue:
719 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
720 // CHECK2:       omp.inner.for.inc:
721 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
722 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
723 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
724 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
725 // CHECK2:       omp.inner.for.end:
726 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
727 // CHECK2:       omp.loop.exit:
728 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
729 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
730 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
731 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
732 // CHECK2:       .omp.final.then:
733 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
734 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
735 // CHECK2:       .omp.final.done:
736 // CHECK2-NEXT:    ret void
737 //
738 //
739 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39
740 // CHECK2-SAME: () #[[ATTR1]] {
741 // CHECK2-NEXT:  entry:
742 // 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*, ...)*))
743 // CHECK2-NEXT:    ret void
744 //
745 //
746 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
747 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
748 // CHECK2-NEXT:  entry:
749 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
750 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
751 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
752 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
753 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
754 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
755 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
756 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
757 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
758 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
759 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
760 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
761 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
762 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
763 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
764 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
765 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
766 // 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)
767 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
768 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
769 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
770 // CHECK2:       cond.true:
771 // CHECK2-NEXT:    br label [[COND_END:%.*]]
772 // CHECK2:       cond.false:
773 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
774 // CHECK2-NEXT:    br label [[COND_END]]
775 // CHECK2:       cond.end:
776 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
777 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
778 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
779 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
780 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
781 // CHECK2:       omp.inner.for.cond:
782 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
783 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
784 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
785 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
786 // CHECK2:       omp.inner.for.body:
787 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
788 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
789 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
790 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
791 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
792 // 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]])
793 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
794 // CHECK2:       omp.inner.for.inc:
795 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
796 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
797 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
798 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
799 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
800 // CHECK2:       omp.inner.for.end:
801 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
802 // CHECK2:       omp.loop.exit:
803 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
804 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
805 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
806 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
807 // CHECK2:       .omp.final.then:
808 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
809 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
810 // CHECK2:       .omp.final.done:
811 // CHECK2-NEXT:    ret void
812 //
813 //
814 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
815 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
816 // CHECK2-NEXT:  entry:
817 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
818 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
819 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
820 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
821 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
822 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
823 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
824 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
825 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
826 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
827 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
828 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
829 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
830 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
831 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
832 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
833 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
834 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
835 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
836 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
837 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
838 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
839 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
840 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
841 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
842 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
843 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
844 // 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)
845 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
846 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
847 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
848 // CHECK2:       cond.true:
849 // CHECK2-NEXT:    br label [[COND_END:%.*]]
850 // CHECK2:       cond.false:
851 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
852 // CHECK2-NEXT:    br label [[COND_END]]
853 // CHECK2:       cond.end:
854 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
855 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
856 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
857 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
858 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
859 // CHECK2:       omp.inner.for.cond:
860 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
861 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
862 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
863 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
864 // CHECK2:       omp.inner.for.body:
865 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
866 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
867 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
868 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
869 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
870 // CHECK2:       omp.body.continue:
871 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
872 // CHECK2:       omp.inner.for.inc:
873 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
874 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
875 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
876 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
877 // CHECK2:       omp.inner.for.end:
878 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
879 // CHECK2:       omp.loop.exit:
880 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
881 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
882 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
883 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
884 // CHECK2:       .omp.final.then:
885 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
886 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
887 // CHECK2:       .omp.final.done:
888 // CHECK2-NEXT:    ret void
889 //
890 //
891 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
892 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat {
893 // CHECK2-NEXT:  entry:
894 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
895 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
896 // 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 0)
897 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
898 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
899 // CHECK2:       omp_offload.failed:
900 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
901 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
902 // CHECK2:       omp_offload.cont:
903 // CHECK2-NEXT:    ret i32 0
904 //
905 //
906 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
907 // CHECK2-SAME: () #[[ATTR1]] {
908 // CHECK2-NEXT:  entry:
909 // 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*, ...)*))
910 // CHECK2-NEXT:    ret void
911 //
912 //
913 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
914 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
915 // CHECK2-NEXT:  entry:
916 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
917 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
918 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
919 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
920 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
921 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
922 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
923 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
924 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
925 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
926 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
927 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
928 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
929 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
930 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
931 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
932 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
933 // 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)
934 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
935 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
936 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
937 // CHECK2:       cond.true:
938 // CHECK2-NEXT:    br label [[COND_END:%.*]]
939 // CHECK2:       cond.false:
940 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
941 // CHECK2-NEXT:    br label [[COND_END]]
942 // CHECK2:       cond.end:
943 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
944 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
945 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
946 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
947 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
948 // CHECK2:       omp.inner.for.cond:
949 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
950 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
951 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
952 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
953 // CHECK2:       omp.inner.for.body:
954 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
955 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
956 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
957 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
958 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
959 // 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]])
960 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
961 // CHECK2:       omp.inner.for.inc:
962 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
963 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
964 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
965 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
966 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
967 // CHECK2:       omp.inner.for.end:
968 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
969 // CHECK2:       omp.loop.exit:
970 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
971 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
972 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
973 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
974 // CHECK2:       .omp.final.then:
975 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
976 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
977 // CHECK2:       .omp.final.done:
978 // CHECK2-NEXT:    ret void
979 //
980 //
981 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
982 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
983 // CHECK2-NEXT:  entry:
984 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
985 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
986 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
987 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
988 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
989 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
990 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
991 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
992 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
993 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
994 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
995 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
996 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
997 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
998 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
999 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1000 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1001 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1002 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1003 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1004 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1005 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1006 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1007 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1008 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1009 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1010 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1011 // 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)
1012 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1013 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
1014 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1015 // CHECK2:       cond.true:
1016 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1017 // CHECK2:       cond.false:
1018 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1019 // CHECK2-NEXT:    br label [[COND_END]]
1020 // CHECK2:       cond.end:
1021 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1022 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1023 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1024 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1025 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1026 // CHECK2:       omp.inner.for.cond:
1027 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1028 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1029 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1030 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1031 // CHECK2:       omp.inner.for.body:
1032 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1033 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1034 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1035 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1036 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1037 // CHECK2:       omp.body.continue:
1038 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1039 // CHECK2:       omp.inner.for.inc:
1040 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1041 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1042 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1043 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1044 // CHECK2:       omp.inner.for.end:
1045 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1046 // CHECK2:       omp.loop.exit:
1047 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1048 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1049 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1050 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1051 // CHECK2:       .omp.final.then:
1052 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
1053 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1054 // CHECK2:       .omp.final.done:
1055 // CHECK2-NEXT:    ret void
1056 //
1057 //
1058 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1059 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
1060 // CHECK2-NEXT:  entry:
1061 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
1062 // CHECK2-NEXT:    ret void
1063 //
1064 //
1065 // CHECK3-LABEL: define {{[^@]+}}@main
1066 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1067 // CHECK3-NEXT:  entry:
1068 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1069 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1072 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1073 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1074 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1075 // CHECK3-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1076 // CHECK3-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1077 // CHECK3-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1078 // CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
1079 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1080 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1081 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1082 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1083 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1084 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1085 // CHECK3:       omp.inner.for.cond:
1086 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1087 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1088 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1089 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1090 // CHECK3:       omp.inner.for.body:
1091 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1092 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1093 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1094 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1095 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1096 // CHECK3:       omp.body.continue:
1097 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1098 // CHECK3:       omp.inner.for.inc:
1099 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1100 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1101 // CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1102 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1103 // CHECK3:       omp.inner.for.end:
1104 // CHECK3-NEXT:    store i32 1000, i32* [[I]], align 4
1105 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
1106 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB4]], align 4
1107 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1108 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1109 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
1110 // CHECK3:       omp.inner.for.cond7:
1111 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1112 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1113 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1114 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1115 // CHECK3:       omp.inner.for.body9:
1116 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1117 // CHECK3-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1118 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1119 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1120 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
1121 // CHECK3:       omp.body.continue12:
1122 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
1123 // CHECK3:       omp.inner.for.inc13:
1124 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1125 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1126 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1127 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1128 // CHECK3:       omp.inner.for.end15:
1129 // CHECK3-NEXT:    store i32 1000, i32* [[I6]], align 4
1130 // CHECK3-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1131 // CHECK3-NEXT:    ret i32 [[CALL]]
1132 //
1133 //
1134 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1135 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] comdat {
1136 // CHECK3-NEXT:  entry:
1137 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1138 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1139 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1140 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1141 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1142 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1143 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1144 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1145 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1146 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1147 // CHECK3:       omp.inner.for.cond:
1148 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1149 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1150 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1151 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1152 // CHECK3:       omp.inner.for.body:
1153 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1154 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1155 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1156 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1157 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1158 // CHECK3:       omp.body.continue:
1159 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1160 // CHECK3:       omp.inner.for.inc:
1161 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1162 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1163 // CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1164 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1165 // CHECK3:       omp.inner.for.end:
1166 // CHECK3-NEXT:    store i32 1000, i32* [[I]], align 4
1167 // CHECK3-NEXT:    ret i32 0
1168 //
1169 //
1170 // CHECK4-LABEL: define {{[^@]+}}@main
1171 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
1172 // CHECK4-NEXT:  entry:
1173 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1174 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1175 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1176 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1177 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1178 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1179 // CHECK4-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1180 // CHECK4-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1181 // CHECK4-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1182 // CHECK4-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1183 // CHECK4-NEXT:    [[I6:%.*]] = alloca i32, align 4
1184 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1185 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1186 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1187 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1188 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1189 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1190 // CHECK4:       omp.inner.for.cond:
1191 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1192 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1193 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1194 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1195 // CHECK4:       omp.inner.for.body:
1196 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1197 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1198 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1199 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1200 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1201 // CHECK4:       omp.body.continue:
1202 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1203 // CHECK4:       omp.inner.for.inc:
1204 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1205 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1206 // CHECK4-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1207 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1208 // CHECK4:       omp.inner.for.end:
1209 // CHECK4-NEXT:    store i32 1000, i32* [[I]], align 4
1210 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
1211 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB4]], align 4
1212 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1213 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1214 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
1215 // CHECK4:       omp.inner.for.cond7:
1216 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1217 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1218 // CHECK4-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1219 // CHECK4-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1220 // CHECK4:       omp.inner.for.body9:
1221 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1222 // CHECK4-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1223 // CHECK4-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1224 // CHECK4-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1225 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
1226 // CHECK4:       omp.body.continue12:
1227 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
1228 // CHECK4:       omp.inner.for.inc13:
1229 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1230 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1231 // CHECK4-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1232 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1233 // CHECK4:       omp.inner.for.end15:
1234 // CHECK4-NEXT:    store i32 1000, i32* [[I6]], align 4
1235 // CHECK4-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1236 // CHECK4-NEXT:    ret i32 [[CALL]]
1237 //
1238 //
1239 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1240 // CHECK4-SAME: () #[[ATTR1:[0-9]+]] comdat {
1241 // CHECK4-NEXT:  entry:
1242 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1243 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1244 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1245 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1246 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1247 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1248 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1249 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1250 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1251 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1252 // CHECK4:       omp.inner.for.cond:
1253 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1254 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1255 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1256 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1257 // CHECK4:       omp.inner.for.body:
1258 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1259 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1260 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1261 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1262 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1263 // CHECK4:       omp.body.continue:
1264 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1265 // CHECK4:       omp.inner.for.inc:
1266 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1267 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1268 // CHECK4-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1269 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1270 // CHECK4:       omp.inner.for.end:
1271 // CHECK4-NEXT:    store i32 1000, i32* [[I]], align 4
1272 // CHECK4-NEXT:    ret i32 0
1273 //
1274