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 
12 // expected-no-diagnostics
13 #ifndef HEADER
14 #define HEADER
15 
16 typedef __INTPTR_TYPE__ intptr_t;
17 
18 
19 void foo();
20 
21 struct S {
22   intptr_t a, b, c;
SS23   S(intptr_t a) : a(a) {}
operator charS24   operator char() { return a; }
~SS25   ~S() {}
26 };
27 
28 template <typename T>
tmain()29 T tmain() {
30 #pragma omp target 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 teams distribute parallel for simd proc_bind(spread)
37   for(int i = 0; i < 1000; i++) {}
38 #pragma omp target teams distribute parallel for simd proc_bind(close)
39   for(int i = 0; i < 1000; i++) {}
40   return tmain<int>();
41 }
42 
43 
44 
45 
46 
47 
48 
49 
50 #endif
51 // CHECK1-LABEL: define {{[^@]+}}@main
52 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
53 // CHECK1-NEXT:  entry:
54 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
55 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
56 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
57 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
58 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
59 // 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)
60 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
61 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
62 // CHECK1:       omp_offload.failed:
63 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
64 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
65 // CHECK1:       omp_offload.cont:
66 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
67 // 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_l38.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
68 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
69 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
70 // CHECK1:       omp_offload.failed2:
71 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38() #[[ATTR2]]
72 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
73 // CHECK1:       omp_offload.cont3:
74 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
75 // CHECK1-NEXT:    ret i32 [[CALL]]
76 //
77 //
78 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
79 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
80 // CHECK1-NEXT:  entry:
81 // 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*, ...)*))
82 // CHECK1-NEXT:    ret void
83 //
84 //
85 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
86 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
87 // CHECK1-NEXT:  entry:
88 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
89 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
90 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
91 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
92 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
93 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
94 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
95 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
98 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
99 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
100 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
101 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
102 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
103 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
104 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
105 // 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)
106 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
107 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
108 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
109 // CHECK1:       cond.true:
110 // CHECK1-NEXT:    br label [[COND_END:%.*]]
111 // CHECK1:       cond.false:
112 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
113 // CHECK1-NEXT:    br label [[COND_END]]
114 // CHECK1:       cond.end:
115 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
116 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
117 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
118 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
119 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
120 // CHECK1:       omp.inner.for.cond:
121 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
122 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
123 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
124 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
125 // CHECK1:       omp.inner.for.body:
126 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
127 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
128 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
129 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
130 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
131 // 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]])
132 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
133 // CHECK1:       omp.inner.for.inc:
134 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
135 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
136 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
137 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
138 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
139 // CHECK1:       omp.inner.for.end:
140 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
141 // CHECK1:       omp.loop.exit:
142 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
143 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
144 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
145 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
146 // CHECK1:       .omp.final.then:
147 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
148 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
149 // CHECK1:       .omp.final.done:
150 // CHECK1-NEXT:    ret void
151 //
152 //
153 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
154 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
155 // CHECK1-NEXT:  entry:
156 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
157 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
158 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
159 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
160 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
163 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
164 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
167 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
168 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
169 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
170 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
171 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
172 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
173 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
174 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
175 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
176 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
177 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
178 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
179 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
180 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
181 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
182 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
183 // 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)
184 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
185 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
186 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
187 // CHECK1:       cond.true:
188 // CHECK1-NEXT:    br label [[COND_END:%.*]]
189 // CHECK1:       cond.false:
190 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
191 // CHECK1-NEXT:    br label [[COND_END]]
192 // CHECK1:       cond.end:
193 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
194 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
195 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
196 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
197 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
198 // CHECK1:       omp.inner.for.cond:
199 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
200 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
201 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
202 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
203 // CHECK1:       omp.inner.for.body:
204 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
205 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
206 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
207 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
208 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
209 // CHECK1:       omp.body.continue:
210 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
211 // CHECK1:       omp.inner.for.inc:
212 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
213 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
214 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
215 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
216 // CHECK1:       omp.inner.for.end:
217 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
218 // CHECK1:       omp.loop.exit:
219 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
220 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
221 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
222 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
223 // CHECK1:       .omp.final.then:
224 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
225 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
226 // CHECK1:       .omp.final.done:
227 // CHECK1-NEXT:    ret void
228 //
229 //
230 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38
231 // CHECK1-SAME: () #[[ATTR1]] {
232 // CHECK1-NEXT:  entry:
233 // 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*, ...)*))
234 // CHECK1-NEXT:    ret void
235 //
236 //
237 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
238 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
239 // CHECK1-NEXT:  entry:
240 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
241 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
242 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
243 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
244 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
245 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
246 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
247 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
248 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
249 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
250 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
251 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
252 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
253 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
254 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
255 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
256 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
257 // 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)
258 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
259 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
260 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
261 // CHECK1:       cond.true:
262 // CHECK1-NEXT:    br label [[COND_END:%.*]]
263 // CHECK1:       cond.false:
264 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
265 // CHECK1-NEXT:    br label [[COND_END]]
266 // CHECK1:       cond.end:
267 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
268 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
269 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
270 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
271 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
272 // CHECK1:       omp.inner.for.cond:
273 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
274 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
275 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
276 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
277 // CHECK1:       omp.inner.for.body:
278 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
279 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
280 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
281 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
282 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
283 // 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]])
284 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
285 // CHECK1:       omp.inner.for.inc:
286 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
287 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
288 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
289 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
290 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
291 // CHECK1:       omp.inner.for.end:
292 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
293 // CHECK1:       omp.loop.exit:
294 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
295 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
296 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
297 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
298 // CHECK1:       .omp.final.then:
299 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
300 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
301 // CHECK1:       .omp.final.done:
302 // CHECK1-NEXT:    ret void
303 //
304 //
305 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
306 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
307 // CHECK1-NEXT:  entry:
308 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
309 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
310 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
311 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
312 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
313 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
314 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
315 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
316 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
317 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
318 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
320 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
321 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
322 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
323 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
324 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
325 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
326 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
327 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
328 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
329 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
330 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
331 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
332 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
333 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
334 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
335 // 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)
336 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
337 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
338 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
339 // CHECK1:       cond.true:
340 // CHECK1-NEXT:    br label [[COND_END:%.*]]
341 // CHECK1:       cond.false:
342 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
343 // CHECK1-NEXT:    br label [[COND_END]]
344 // CHECK1:       cond.end:
345 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
346 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
347 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
348 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
349 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
350 // CHECK1:       omp.inner.for.cond:
351 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
352 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
353 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
354 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
355 // CHECK1:       omp.inner.for.body:
356 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
357 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
358 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
359 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
360 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
361 // CHECK1:       omp.body.continue:
362 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
363 // CHECK1:       omp.inner.for.inc:
364 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
365 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
366 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
367 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
368 // CHECK1:       omp.inner.for.end:
369 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
370 // CHECK1:       omp.loop.exit:
371 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
372 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
373 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
374 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
375 // CHECK1:       .omp.final.then:
376 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
377 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
378 // CHECK1:       .omp.final.done:
379 // CHECK1-NEXT:    ret void
380 //
381 //
382 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
383 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
384 // CHECK1-NEXT:  entry:
385 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
387 // 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_l30.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
388 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
389 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
390 // CHECK1:       omp_offload.failed:
391 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30() #[[ATTR2]]
392 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
393 // CHECK1:       omp_offload.cont:
394 // CHECK1-NEXT:    ret i32 0
395 //
396 //
397 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30
398 // CHECK1-SAME: () #[[ATTR1]] {
399 // CHECK1-NEXT:  entry:
400 // 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*, ...)*))
401 // CHECK1-NEXT:    ret void
402 //
403 //
404 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
405 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
406 // CHECK1-NEXT:  entry:
407 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
408 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
409 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
410 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
411 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
412 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
413 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
414 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
415 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
416 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
417 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
418 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
419 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
420 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
421 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
422 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
423 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
424 // 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)
425 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
426 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
427 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
428 // CHECK1:       cond.true:
429 // CHECK1-NEXT:    br label [[COND_END:%.*]]
430 // CHECK1:       cond.false:
431 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
432 // CHECK1-NEXT:    br label [[COND_END]]
433 // CHECK1:       cond.end:
434 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
435 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
436 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
437 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
438 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
439 // CHECK1:       omp.inner.for.cond:
440 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
441 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
442 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
443 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
444 // CHECK1:       omp.inner.for.body:
445 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
446 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
447 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
448 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
449 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
450 // 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]])
451 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
452 // CHECK1:       omp.inner.for.inc:
453 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
454 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
455 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
456 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
457 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
458 // CHECK1:       omp.inner.for.end:
459 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
460 // CHECK1:       omp.loop.exit:
461 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
462 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
463 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
464 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
465 // CHECK1:       .omp.final.then:
466 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
467 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
468 // CHECK1:       .omp.final.done:
469 // CHECK1-NEXT:    ret void
470 //
471 //
472 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
473 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
474 // CHECK1-NEXT:  entry:
475 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
476 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
477 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
478 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
479 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
480 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
481 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
482 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
483 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
484 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
485 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
486 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
487 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
488 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
489 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
490 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
491 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
492 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
493 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
494 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
495 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
496 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
497 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
498 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
499 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
500 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
501 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
502 // 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)
503 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
504 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
505 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
506 // CHECK1:       cond.true:
507 // CHECK1-NEXT:    br label [[COND_END:%.*]]
508 // CHECK1:       cond.false:
509 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
510 // CHECK1-NEXT:    br label [[COND_END]]
511 // CHECK1:       cond.end:
512 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
513 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
514 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
515 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
516 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
517 // CHECK1:       omp.inner.for.cond:
518 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
519 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
520 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
521 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
522 // CHECK1:       omp.inner.for.body:
523 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
524 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
525 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
526 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
527 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
528 // CHECK1:       omp.body.continue:
529 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
530 // CHECK1:       omp.inner.for.inc:
531 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
532 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
533 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
534 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
535 // CHECK1:       omp.inner.for.end:
536 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
537 // CHECK1:       omp.loop.exit:
538 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
539 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
540 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
541 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
542 // CHECK1:       .omp.final.then:
543 // CHECK1-NEXT:    store i32 1000, i32* [[I]], align 4
544 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
545 // CHECK1:       .omp.final.done:
546 // CHECK1-NEXT:    ret void
547 //
548 //
549 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
550 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
551 // CHECK1-NEXT:  entry:
552 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
553 // CHECK1-NEXT:    ret void
554 //
555 //
556 // CHECK2-LABEL: define {{[^@]+}}@main
557 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
558 // CHECK2-NEXT:  entry:
559 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
560 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
561 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
562 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
563 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
564 // 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)
565 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
566 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
567 // CHECK2:       omp_offload.failed:
568 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
569 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
570 // CHECK2:       omp_offload.cont:
571 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
572 // 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_l38.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
573 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
574 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
575 // CHECK2:       omp_offload.failed2:
576 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38() #[[ATTR2]]
577 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
578 // CHECK2:       omp_offload.cont3:
579 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
580 // CHECK2-NEXT:    ret i32 [[CALL]]
581 //
582 //
583 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
584 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
585 // CHECK2-NEXT:  entry:
586 // 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*, ...)*))
587 // CHECK2-NEXT:    ret void
588 //
589 //
590 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
591 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
592 // CHECK2-NEXT:  entry:
593 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
594 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
595 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
596 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
597 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
598 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
599 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
600 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
601 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
602 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
603 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
604 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
605 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
606 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
607 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
608 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
609 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
610 // 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)
611 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
612 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
613 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
614 // CHECK2:       cond.true:
615 // CHECK2-NEXT:    br label [[COND_END:%.*]]
616 // CHECK2:       cond.false:
617 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
618 // CHECK2-NEXT:    br label [[COND_END]]
619 // CHECK2:       cond.end:
620 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
621 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
622 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
623 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
624 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
625 // CHECK2:       omp.inner.for.cond:
626 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
627 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
628 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
629 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
630 // CHECK2:       omp.inner.for.body:
631 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
632 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
633 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
634 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
635 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
636 // 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]])
637 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
638 // CHECK2:       omp.inner.for.inc:
639 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
640 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
641 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
642 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
643 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
644 // CHECK2:       omp.inner.for.end:
645 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
646 // CHECK2:       omp.loop.exit:
647 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
648 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
649 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
650 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
651 // CHECK2:       .omp.final.then:
652 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
653 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
654 // CHECK2:       .omp.final.done:
655 // CHECK2-NEXT:    ret void
656 //
657 //
658 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
659 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
660 // CHECK2-NEXT:  entry:
661 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
662 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
663 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
664 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
665 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
666 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
667 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
668 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
669 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
670 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
671 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
672 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
673 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
674 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
675 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
676 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
677 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
678 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
679 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
680 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
681 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
682 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
683 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
684 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
685 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
686 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
687 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
688 // 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)
689 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
690 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
691 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
692 // CHECK2:       cond.true:
693 // CHECK2-NEXT:    br label [[COND_END:%.*]]
694 // CHECK2:       cond.false:
695 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
696 // CHECK2-NEXT:    br label [[COND_END]]
697 // CHECK2:       cond.end:
698 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
699 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
700 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
701 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
702 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
703 // CHECK2:       omp.inner.for.cond:
704 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
705 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
706 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
707 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
708 // CHECK2:       omp.inner.for.body:
709 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
710 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
711 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
712 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
713 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
714 // CHECK2:       omp.body.continue:
715 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
716 // CHECK2:       omp.inner.for.inc:
717 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
718 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
719 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
720 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
721 // CHECK2:       omp.inner.for.end:
722 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
723 // CHECK2:       omp.loop.exit:
724 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
725 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
726 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
727 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
728 // CHECK2:       .omp.final.then:
729 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
730 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
731 // CHECK2:       .omp.final.done:
732 // CHECK2-NEXT:    ret void
733 //
734 //
735 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38
736 // CHECK2-SAME: () #[[ATTR1]] {
737 // CHECK2-NEXT:  entry:
738 // 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*, ...)*))
739 // CHECK2-NEXT:    ret void
740 //
741 //
742 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
743 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
744 // CHECK2-NEXT:  entry:
745 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
746 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
747 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
748 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
749 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
750 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
751 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
752 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
753 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
754 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
755 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
756 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
757 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
758 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
759 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
760 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
761 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
762 // 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)
763 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
764 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
765 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
766 // CHECK2:       cond.true:
767 // CHECK2-NEXT:    br label [[COND_END:%.*]]
768 // CHECK2:       cond.false:
769 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
770 // CHECK2-NEXT:    br label [[COND_END]]
771 // CHECK2:       cond.end:
772 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
773 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
774 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
775 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
776 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
777 // CHECK2:       omp.inner.for.cond:
778 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
779 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
780 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
781 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
782 // CHECK2:       omp.inner.for.body:
783 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
784 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
785 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
786 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
787 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
788 // 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]])
789 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
790 // CHECK2:       omp.inner.for.inc:
791 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
792 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
793 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
794 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
795 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
796 // CHECK2:       omp.inner.for.end:
797 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
798 // CHECK2:       omp.loop.exit:
799 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
800 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
801 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
802 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
803 // CHECK2:       .omp.final.then:
804 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
805 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
806 // CHECK2:       .omp.final.done:
807 // CHECK2-NEXT:    ret void
808 //
809 //
810 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
811 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
812 // CHECK2-NEXT:  entry:
813 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
814 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
815 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
816 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
817 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
818 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
819 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
820 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
821 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
822 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
823 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
824 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
825 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
826 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
827 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
828 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
829 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
830 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
831 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
832 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
833 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
834 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
835 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
836 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
837 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
838 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
839 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
840 // 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)
841 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
842 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
843 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
844 // CHECK2:       cond.true:
845 // CHECK2-NEXT:    br label [[COND_END:%.*]]
846 // CHECK2:       cond.false:
847 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
848 // CHECK2-NEXT:    br label [[COND_END]]
849 // CHECK2:       cond.end:
850 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
851 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
852 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
853 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
854 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
855 // CHECK2:       omp.inner.for.cond:
856 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
857 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
858 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
859 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
860 // CHECK2:       omp.inner.for.body:
861 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
862 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
863 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
864 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
865 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
866 // CHECK2:       omp.body.continue:
867 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
868 // CHECK2:       omp.inner.for.inc:
869 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
870 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
871 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
872 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
873 // CHECK2:       omp.inner.for.end:
874 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
875 // CHECK2:       omp.loop.exit:
876 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
877 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
878 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
879 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
880 // CHECK2:       .omp.final.then:
881 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
882 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
883 // CHECK2:       .omp.final.done:
884 // CHECK2-NEXT:    ret void
885 //
886 //
887 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
888 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat {
889 // CHECK2-NEXT:  entry:
890 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
891 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
892 // 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_l30.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
893 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
894 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
895 // CHECK2:       omp_offload.failed:
896 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30() #[[ATTR2]]
897 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
898 // CHECK2:       omp_offload.cont:
899 // CHECK2-NEXT:    ret i32 0
900 //
901 //
902 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30
903 // CHECK2-SAME: () #[[ATTR1]] {
904 // CHECK2-NEXT:  entry:
905 // 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*, ...)*))
906 // CHECK2-NEXT:    ret void
907 //
908 //
909 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
910 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
911 // CHECK2-NEXT:  entry:
912 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
913 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
914 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
915 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
916 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
917 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
918 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
919 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
920 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
921 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
922 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
923 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
924 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
925 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
926 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
927 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
928 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
929 // 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)
930 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
931 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
932 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
933 // CHECK2:       cond.true:
934 // CHECK2-NEXT:    br label [[COND_END:%.*]]
935 // CHECK2:       cond.false:
936 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
937 // CHECK2-NEXT:    br label [[COND_END]]
938 // CHECK2:       cond.end:
939 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
940 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
941 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
942 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
943 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
944 // CHECK2:       omp.inner.for.cond:
945 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
946 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
947 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
948 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
949 // CHECK2:       omp.inner.for.body:
950 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
951 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
952 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
953 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
954 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
955 // 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]])
956 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
957 // CHECK2:       omp.inner.for.inc:
958 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
959 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
960 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
961 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
962 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
963 // CHECK2:       omp.inner.for.end:
964 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
965 // CHECK2:       omp.loop.exit:
966 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
967 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
968 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
969 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
970 // CHECK2:       .omp.final.then:
971 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
972 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
973 // CHECK2:       .omp.final.done:
974 // CHECK2-NEXT:    ret void
975 //
976 //
977 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
978 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
979 // CHECK2-NEXT:  entry:
980 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
981 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
982 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
983 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
984 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
985 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
986 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
987 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
988 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
989 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
990 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
991 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
992 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
993 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
994 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
995 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
996 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
997 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
998 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
999 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1000 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1001 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1002 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1003 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1004 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1005 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1006 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1007 // 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)
1008 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1009 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
1010 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1011 // CHECK2:       cond.true:
1012 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1013 // CHECK2:       cond.false:
1014 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1015 // CHECK2-NEXT:    br label [[COND_END]]
1016 // CHECK2:       cond.end:
1017 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1018 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1019 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1020 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1021 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1022 // CHECK2:       omp.inner.for.cond:
1023 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1024 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1025 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1026 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1027 // CHECK2:       omp.inner.for.body:
1028 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1029 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1030 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1031 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1032 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1033 // CHECK2:       omp.body.continue:
1034 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1035 // CHECK2:       omp.inner.for.inc:
1036 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1037 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1038 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1039 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1040 // CHECK2:       omp.inner.for.end:
1041 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1042 // CHECK2:       omp.loop.exit:
1043 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1044 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1045 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1046 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1047 // CHECK2:       .omp.final.then:
1048 // CHECK2-NEXT:    store i32 1000, i32* [[I]], align 4
1049 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1050 // CHECK2:       .omp.final.done:
1051 // CHECK2-NEXT:    ret void
1052 //
1053 //
1054 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1055 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
1056 // CHECK2-NEXT:  entry:
1057 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
1058 // CHECK2-NEXT:    ret void
1059 //
1060 //
1061 // CHECK3-LABEL: define {{[^@]+}}@main
1062 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1063 // CHECK3-NEXT:  entry:
1064 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1065 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1066 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1067 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1068 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1069 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1072 // CHECK3-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1073 // CHECK3-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1074 // CHECK3-NEXT:    [[I6:%.*]] = alloca i32, align 4
1075 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1076 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1077 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1078 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1079 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1080 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1081 // CHECK3:       omp.inner.for.cond:
1082 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1083 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1084 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1085 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1086 // CHECK3:       omp.inner.for.body:
1087 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1088 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1089 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1090 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1091 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1092 // CHECK3:       omp.body.continue:
1093 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1094 // CHECK3:       omp.inner.for.inc:
1095 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1096 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1097 // CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1098 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1099 // CHECK3:       omp.inner.for.end:
1100 // CHECK3-NEXT:    store i32 1000, i32* [[I]], align 4
1101 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
1102 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB4]], align 4
1103 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1104 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1105 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
1106 // CHECK3:       omp.inner.for.cond7:
1107 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1108 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1109 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1110 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1111 // CHECK3:       omp.inner.for.body9:
1112 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1113 // CHECK3-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1114 // CHECK3-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1115 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1116 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
1117 // CHECK3:       omp.body.continue12:
1118 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
1119 // CHECK3:       omp.inner.for.inc13:
1120 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1121 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1122 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1123 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1124 // CHECK3:       omp.inner.for.end15:
1125 // CHECK3-NEXT:    store i32 1000, i32* [[I6]], align 4
1126 // CHECK3-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1127 // CHECK3-NEXT:    ret i32 [[CALL]]
1128 //
1129 //
1130 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1131 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] comdat {
1132 // CHECK3-NEXT:  entry:
1133 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1134 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1135 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1136 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1137 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1138 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1139 // CHECK3-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1140 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1141 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1142 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1143 // CHECK3:       omp.inner.for.cond:
1144 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1145 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1146 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1147 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1148 // CHECK3:       omp.inner.for.body:
1149 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1150 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1151 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1152 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1153 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1154 // CHECK3:       omp.body.continue:
1155 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1156 // CHECK3:       omp.inner.for.inc:
1157 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1158 // CHECK3-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1159 // CHECK3-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1160 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1161 // CHECK3:       omp.inner.for.end:
1162 // CHECK3-NEXT:    store i32 1000, i32* [[I]], align 4
1163 // CHECK3-NEXT:    ret i32 0
1164 //
1165 //
1166 // CHECK4-LABEL: define {{[^@]+}}@main
1167 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
1168 // CHECK4-NEXT:  entry:
1169 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1170 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1171 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1172 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1173 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1174 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1175 // CHECK4-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1176 // CHECK4-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
1177 // CHECK4-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
1178 // CHECK4-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
1179 // CHECK4-NEXT:    [[I6:%.*]] = alloca i32, align 4
1180 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1181 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1182 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1183 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1184 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1185 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1186 // CHECK4:       omp.inner.for.cond:
1187 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1188 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
1189 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1190 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1191 // CHECK4:       omp.inner.for.body:
1192 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1193 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1194 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1195 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
1196 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1197 // CHECK4:       omp.body.continue:
1198 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1199 // CHECK4:       omp.inner.for.inc:
1200 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1201 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1202 // CHECK4-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
1203 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
1204 // CHECK4:       omp.inner.for.end:
1205 // CHECK4-NEXT:    store i32 1000, i32* [[I]], align 4
1206 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
1207 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB4]], align 4
1208 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
1209 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
1210 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
1211 // CHECK4:       omp.inner.for.cond7:
1212 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1213 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
1214 // CHECK4-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1215 // CHECK4-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
1216 // CHECK4:       omp.inner.for.body9:
1217 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1218 // CHECK4-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
1219 // CHECK4-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
1220 // CHECK4-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
1221 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
1222 // CHECK4:       omp.body.continue12:
1223 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
1224 // CHECK4:       omp.inner.for.inc13:
1225 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1226 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
1227 // CHECK4-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
1228 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
1229 // CHECK4:       omp.inner.for.end15:
1230 // CHECK4-NEXT:    store i32 1000, i32* [[I6]], align 4
1231 // CHECK4-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
1232 // CHECK4-NEXT:    ret i32 [[CALL]]
1233 //
1234 //
1235 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1236 // CHECK4-SAME: () #[[ATTR1:[0-9]+]] comdat {
1237 // CHECK4-NEXT:  entry:
1238 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1239 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1240 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1241 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1242 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
1243 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1244 // CHECK4-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
1245 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1246 // CHECK4-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
1247 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1248 // CHECK4:       omp.inner.for.cond:
1249 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1250 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
1251 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
1252 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1253 // CHECK4:       omp.inner.for.body:
1254 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1255 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
1256 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1257 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
1258 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1259 // CHECK4:       omp.body.continue:
1260 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1261 // CHECK4:       omp.inner.for.inc:
1262 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1263 // CHECK4-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
1264 // CHECK4-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
1265 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1266 // CHECK4:       omp.inner.for.end:
1267 // CHECK4-NEXT:    store i32 1000, i32* [[I]], align 4
1268 // CHECK4-NEXT:    ret i32 0
1269 //
1270