1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
6 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK4
7 
8 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
12 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 // expected-no-diagnostics
14 #ifndef HEADER
15 #define HEADER
16 
17 void foo(float *c) {
18 #pragma omp parallel for simd aligned(c)
19   for (int i = 0; i < 10; ++i);
20 }
21 
22 #endif
23 
24 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooPf
25 // CHECK1-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
26 // CHECK1-NEXT:  entry:
27 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
28 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
29 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
30 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
31 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP0]], align 8
32 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
33 // CHECK1-NEXT:    ret void
34 //
35 //
36 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
37 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
38 // CHECK1-NEXT:  entry:
39 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
40 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
41 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
42 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
43 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
44 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
45 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
46 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
47 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
48 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
49 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
50 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
51 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
52 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
53 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
54 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
55 // CHECK1-NEXT:    [[TMP3:%.*]] = load float*, float** [[TMP2]], align 8
56 // CHECK1-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP3]], i64 16) ]
57 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
58 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
59 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
60 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
61 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
62 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
63 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
64 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
65 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
66 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
67 // CHECK1:       cond.true:
68 // CHECK1-NEXT:    br label [[COND_END:%.*]]
69 // CHECK1:       cond.false:
70 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
71 // CHECK1-NEXT:    br label [[COND_END]]
72 // CHECK1:       cond.end:
73 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
74 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
75 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
76 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
77 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
78 // CHECK1:       omp.inner.for.cond:
79 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
80 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
81 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
82 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
83 // CHECK1:       omp.inner.for.body:
84 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
85 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
86 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
87 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
88 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
89 // CHECK1:       omp.body.continue:
90 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
91 // CHECK1:       omp.inner.for.inc:
92 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
93 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
94 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
95 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
96 // CHECK1:       omp.inner.for.end:
97 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
98 // CHECK1:       omp.loop.exit:
99 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
100 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
101 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
102 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
103 // CHECK1:       .omp.final.then:
104 // CHECK1-NEXT:    store i32 10, i32* [[I]], align 4
105 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
106 // CHECK1:       .omp.final.done:
107 // CHECK1-NEXT:    ret void
108 //
109 //
110 // CHECK2-LABEL: define {{[^@]+}}@_Z3fooPf
111 // CHECK2-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
112 // CHECK2-NEXT:  entry:
113 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
114 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
115 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
116 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
117 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP0]], align 8
118 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
119 // CHECK2-NEXT:    ret void
120 //
121 //
122 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
123 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
124 // CHECK2-NEXT:  entry:
125 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
126 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
127 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
128 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
129 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
130 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
131 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
132 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
133 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
134 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
135 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
136 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
137 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
138 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
139 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
140 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
141 // CHECK2-NEXT:    [[TMP3:%.*]] = load float*, float** [[TMP2]], align 8
142 // CHECK2-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP3]], i64 16) ]
143 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
144 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
145 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
146 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
147 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
148 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
149 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
150 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
151 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
152 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
153 // CHECK2:       cond.true:
154 // CHECK2-NEXT:    br label [[COND_END:%.*]]
155 // CHECK2:       cond.false:
156 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
157 // CHECK2-NEXT:    br label [[COND_END]]
158 // CHECK2:       cond.end:
159 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
160 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
161 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
162 // CHECK2-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
163 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
164 // CHECK2:       omp.inner.for.cond:
165 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
166 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
167 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
168 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
169 // CHECK2:       omp.inner.for.body:
170 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
171 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
172 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
173 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
174 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
175 // CHECK2:       omp.body.continue:
176 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
177 // CHECK2:       omp.inner.for.inc:
178 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
179 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
180 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
181 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
182 // CHECK2:       omp.inner.for.end:
183 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
184 // CHECK2:       omp.loop.exit:
185 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
186 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
187 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
188 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
189 // CHECK2:       .omp.final.then:
190 // CHECK2-NEXT:    store i32 10, i32* [[I]], align 4
191 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
192 // CHECK2:       .omp.final.done:
193 // CHECK2-NEXT:    ret void
194 //
195 //
196 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooPf
197 // CHECK3-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
198 // CHECK3-NEXT:  entry:
199 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
200 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
201 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
202 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
203 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP0]], align 8
204 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
205 // CHECK3-NEXT:    ret void
206 //
207 //
208 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
209 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
210 // CHECK3-NEXT:  entry:
211 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
212 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
213 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
214 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
215 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
216 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
217 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
218 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
219 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
220 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
221 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
222 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
223 // CHECK3-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
224 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
225 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
226 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
227 // CHECK3-NEXT:    [[TMP3:%.*]] = load float*, float** [[TMP2]], align 8
228 // CHECK3-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP3]], i64 16) ]
229 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
230 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
231 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
232 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
233 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
234 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
235 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
236 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
237 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
238 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
239 // CHECK3:       cond.true:
240 // CHECK3-NEXT:    br label [[COND_END:%.*]]
241 // CHECK3:       cond.false:
242 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
243 // CHECK3-NEXT:    br label [[COND_END]]
244 // CHECK3:       cond.end:
245 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
246 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
247 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
248 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
249 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
250 // CHECK3:       omp.inner.for.cond:
251 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
252 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
253 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
254 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
255 // CHECK3:       omp.inner.for.body:
256 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
257 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
258 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
259 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
260 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
261 // CHECK3:       omp.body.continue:
262 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
263 // CHECK3:       omp.inner.for.inc:
264 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
265 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
266 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
267 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
268 // CHECK3:       omp.inner.for.end:
269 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
270 // CHECK3:       omp.loop.exit:
271 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
272 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
273 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
274 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
275 // CHECK3:       .omp.final.then:
276 // CHECK3-NEXT:    store i32 10, i32* [[I]], align 4
277 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
278 // CHECK3:       .omp.final.done:
279 // CHECK3-NEXT:    ret void
280 //
281 //
282 // CHECK4-LABEL: define {{[^@]+}}@_Z3fooPf
283 // CHECK4-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
284 // CHECK4-NEXT:  entry:
285 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
286 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
287 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
288 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
289 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP0]], align 8
290 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
291 // CHECK4-NEXT:    ret void
292 //
293 //
294 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
295 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
296 // CHECK4-NEXT:  entry:
297 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
298 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
299 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
300 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
301 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
302 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
303 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
304 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
305 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
306 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
307 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
308 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
309 // CHECK4-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
310 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
311 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
312 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
313 // CHECK4-NEXT:    [[TMP3:%.*]] = load float*, float** [[TMP2]], align 8
314 // CHECK4-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP3]], i64 16) ]
315 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
316 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
317 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
318 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
319 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
320 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
321 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
322 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
323 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 9
324 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
325 // CHECK4:       cond.true:
326 // CHECK4-NEXT:    br label [[COND_END:%.*]]
327 // CHECK4:       cond.false:
328 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
329 // CHECK4-NEXT:    br label [[COND_END]]
330 // CHECK4:       cond.end:
331 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
332 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
333 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
334 // CHECK4-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
335 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
336 // CHECK4:       omp.inner.for.cond:
337 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
338 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
339 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
340 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
341 // CHECK4:       omp.inner.for.body:
342 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
343 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
344 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
345 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
346 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
347 // CHECK4:       omp.body.continue:
348 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
349 // CHECK4:       omp.inner.for.inc:
350 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
351 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
352 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
353 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
354 // CHECK4:       omp.inner.for.end:
355 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
356 // CHECK4:       omp.loop.exit:
357 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
358 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
359 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
360 // CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
361 // CHECK4:       .omp.final.then:
362 // CHECK4-NEXT:    store i32 10, i32* [[I]], align 4
363 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
364 // CHECK4:       .omp.final.done:
365 // CHECK4-NEXT:    ret void
366 //
367