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 -DOMP5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5 
6 // RUN: %clang_cc1 -verify -fopenmp-simd -DOMP5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // expected-no-diagnostics
10 
11 #ifndef HEADER
12 #define HEADER
13 
main()14 int main() {
15   int a = 0;
16 #pragma omp parallel for lastprivate(conditional: a)
17   for (int i = 0; i < 10; ++i) {
18     if (i < 5) {
19       a = 0;
20 #pragma omp parallel reduction(+:a) num_threads(10)
21       a += i;
22 #pragma omp atomic
23       a += i;
24 #pragma omp parallel num_threads(10)
25 #pragma omp atomic
26       a += i;
27     }
28   }
29   return 0;
30 }
31 
32 
33 
34 
35 #endif // HEADER
36 // CHECK1-LABEL: define {{[^@]+}}@main
37 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
38 // CHECK1-NEXT:  entry:
39 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
40 // CHECK1-NEXT:    [[A:%.*]] = alloca i32, align 4
41 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
42 // CHECK1-NEXT:    store i32 0, i32* [[A]], align 4
43 // 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*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[A]])
44 // CHECK1-NEXT:    ret i32 0
45 //
46 //
47 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
48 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
49 // CHECK1-NEXT:  entry:
50 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
51 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
52 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
53 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
54 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
55 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
56 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
57 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
58 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
59 // CHECK1-NEXT:    [[A1:%.*]] = alloca [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], align 4
60 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
61 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
62 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
63 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
64 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
65 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
66 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
67 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
68 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
69 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 1
70 // CHECK1-NEXT:    store i8 0, i8* [[TMP1]], align 4
71 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 0
72 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
73 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
74 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
75 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
76 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
77 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
78 // CHECK1:       cond.true:
79 // CHECK1-NEXT:    br label [[COND_END:%.*]]
80 // CHECK1:       cond.false:
81 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
82 // CHECK1-NEXT:    br label [[COND_END]]
83 // CHECK1:       cond.end:
84 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
85 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
86 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
87 // CHECK1-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
88 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
89 // CHECK1:       omp.inner.for.cond:
90 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
91 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
92 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
93 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
94 // CHECK1:       omp.inner.for.body:
95 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
96 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
97 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
98 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
99 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
100 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp slt i32 [[TMP11]], 5
101 // CHECK1-NEXT:    br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
102 // CHECK1:       if.then:
103 // CHECK1-NEXT:    store i32 0, i32* [[TMP2]], align 4
104 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
105 // CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
106 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
107 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp sle i32 [[TMP13]], [[TMP12]]
108 // CHECK1-NEXT:    br i1 [[TMP14]], label [[LP_COND_THEN:%.*]], label [[LP_COND_EXIT:%.*]]
109 // CHECK1:       lp_cond_then:
110 // CHECK1-NEXT:    store i32 [[TMP12]], i32* @.{{pl_cond[.].+[.|,]}} align 4
111 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP2]], align 4
112 // CHECK1-NEXT:    store i32 [[TMP15]], i32* @{{pl_cond[.].+[.|,]}} align 4
113 // CHECK1-NEXT:    br label [[LP_COND_EXIT]]
114 // CHECK1:       lp_cond_exit:
115 // CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
116 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 10)
117 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[TMP2]], i32* [[I]])
118 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
119 // CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
120 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
121 // CHECK1-NEXT:    [[TMP18:%.*]] = icmp sle i32 [[TMP17]], [[TMP16]]
122 // CHECK1-NEXT:    br i1 [[TMP18]], label [[LP_COND_THEN4:%.*]], label [[LP_COND_EXIT5:%.*]]
123 // CHECK1:       lp_cond_then4:
124 // CHECK1-NEXT:    store i32 [[TMP16]], i32* @.{{pl_cond[.].+[.|,]}} align 4
125 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP2]], align 4
126 // CHECK1-NEXT:    store i32 [[TMP19]], i32* @{{pl_cond[.].+[.|,]}} align 4
127 // CHECK1-NEXT:    br label [[LP_COND_EXIT5]]
128 // CHECK1:       lp_cond_exit5:
129 // CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
130 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
131 // CHECK1-NEXT:    [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4
132 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
133 // CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
134 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
135 // CHECK1-NEXT:    [[TMP24:%.*]] = icmp sle i32 [[TMP23]], [[TMP22]]
136 // CHECK1-NEXT:    br i1 [[TMP24]], label [[LP_COND_THEN6:%.*]], label [[LP_COND_EXIT7:%.*]]
137 // CHECK1:       lp_cond_then6:
138 // CHECK1-NEXT:    store i32 [[TMP22]], i32* @.{{pl_cond[.].+[.|,]}} align 4
139 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP2]], align 4
140 // CHECK1-NEXT:    store i32 [[TMP25]], i32* @{{pl_cond[.].+[.|,]}} align 4
141 // CHECK1-NEXT:    br label [[LP_COND_EXIT7]]
142 // CHECK1:       lp_cond_exit7:
143 // CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
144 // CHECK1-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 10)
145 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[TMP2]], i32* [[I]])
146 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 1
147 // CHECK1-NEXT:    [[TMP27:%.*]] = load i8, i8* [[TMP26]], align 4
148 // CHECK1-NEXT:    [[TMP28:%.*]] = icmp ne i8 [[TMP27]], 0
149 // CHECK1-NEXT:    br i1 [[TMP28]], label [[LPC_THEN:%.*]], label [[LPC_DONE:%.*]]
150 // CHECK1:       lpc.then:
151 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
152 // CHECK1-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
153 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
154 // CHECK1-NEXT:    [[TMP31:%.*]] = icmp sle i32 [[TMP30]], [[TMP29]]
155 // CHECK1-NEXT:    br i1 [[TMP31]], label [[LP_COND_THEN8:%.*]], label [[LP_COND_EXIT9:%.*]]
156 // CHECK1:       lp_cond_then8:
157 // CHECK1-NEXT:    store i32 [[TMP29]], i32* @.{{pl_cond[.].+[.|,]}} align 4
158 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP2]], align 4
159 // CHECK1-NEXT:    store i32 [[TMP32]], i32* @{{pl_cond[.].+[.|,]}} align 4
160 // CHECK1-NEXT:    br label [[LP_COND_EXIT9]]
161 // CHECK1:       lp_cond_exit9:
162 // CHECK1-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
163 // CHECK1-NEXT:    br label [[LPC_DONE]]
164 // CHECK1:       lpc.done:
165 // CHECK1-NEXT:    br label [[IF_END]]
166 // CHECK1:       if.end:
167 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
168 // CHECK1:       omp.body.continue:
169 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
170 // CHECK1:       omp.inner.for.inc:
171 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
172 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP33]], 1
173 // CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
174 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
175 // CHECK1:       omp.inner.for.end:
176 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
177 // CHECK1:       omp.loop.exit:
178 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
179 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
180 // CHECK1-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
181 // CHECK1-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP4]])
182 // CHECK1-NEXT:    br i1 [[TMP35]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
183 // CHECK1:       .omp.lastprivate.then:
184 // CHECK1-NEXT:    [[TMP36:%.*]] = load i32, i32* @{{pl_cond[.].+[.|,]}} align 4
185 // CHECK1-NEXT:    store i32 [[TMP36]], i32* [[TMP2]], align 4
186 // CHECK1-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP2]], align 4
187 // CHECK1-NEXT:    store i32 [[TMP37]], i32* [[TMP0]], align 4
188 // CHECK1-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
189 // CHECK1:       .omp.lastprivate.done:
190 // CHECK1-NEXT:    ret void
191 //
192 //
193 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
194 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32* nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
195 // CHECK1-NEXT:  entry:
196 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
197 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
198 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
199 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
200 // CHECK1-NEXT:    [[A1:%.*]] = alloca i32, align 4
201 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
202 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
203 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
204 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
205 // CHECK1-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
206 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
207 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[I_ADDR]], align 8
208 // CHECK1-NEXT:    store i32 0, i32* [[A1]], align 4
209 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
210 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A1]], align 4
211 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], [[TMP2]]
212 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[A1]], align 4
213 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
214 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i32* [[A1]] to i8*
215 // CHECK1-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 8
216 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
217 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
218 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
219 // CHECK1-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 1, i64 8, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
220 // CHECK1-NEXT:    switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
221 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
222 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
223 // CHECK1-NEXT:    ]
224 // CHECK1:       .omp.reduction.case1:
225 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
226 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A1]], align 4
227 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
228 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[TMP0]], align 4
229 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
230 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
231 // CHECK1:       .omp.reduction.case2:
232 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[A1]], align 4
233 // CHECK1-NEXT:    [[TMP13:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP12]] monotonic, align 4
234 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
235 // CHECK1:       .omp.reduction.default:
236 // CHECK1-NEXT:    ret void
237 //
238 //
239 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
240 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
241 // CHECK1-NEXT:  entry:
242 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
243 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
244 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
245 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
246 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
247 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
248 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
249 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
250 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
251 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
252 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
253 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
254 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
255 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
256 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
257 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
258 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
259 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
260 // CHECK1-NEXT:    ret void
261 //
262 //
263 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
264 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32* nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
265 // CHECK1-NEXT:  entry:
266 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
267 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
268 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
269 // CHECK1-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
270 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
271 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
272 // CHECK1-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
273 // CHECK1-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
274 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
275 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[I_ADDR]], align 8
276 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
277 // CHECK1-NEXT:    [[TMP3:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP2]] monotonic, align 4
278 // CHECK1-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to %struct.lasprivate.conditional*
279 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], %struct.lasprivate.conditional* [[TMP4]], i32 0, i32 1
280 // CHECK1-NEXT:    store atomic volatile i8 1, i8* [[TMP5]] unordered, align 1
281 // CHECK1-NEXT:    ret void
282 //
283 //
284 // CHECK2-LABEL: define {{[^@]+}}@main
285 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
286 // CHECK2-NEXT:  entry:
287 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
288 // CHECK2-NEXT:    [[A:%.*]] = alloca i32, align 4
289 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
290 // CHECK2-NEXT:    store i32 0, i32* [[A]], align 4
291 // 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*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[A]])
292 // CHECK2-NEXT:    ret i32 0
293 //
294 //
295 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
296 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
297 // CHECK2-NEXT:  entry:
298 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
299 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
300 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
301 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
302 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
303 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
304 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
305 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
306 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
307 // CHECK2-NEXT:    [[A1:%.*]] = alloca [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], align 4
308 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
309 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
310 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
311 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
312 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
313 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
314 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
315 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
316 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
317 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 1
318 // CHECK2-NEXT:    store i8 0, i8* [[TMP1]], align 4
319 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 0
320 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
321 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
322 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
323 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
324 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
325 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
326 // CHECK2:       cond.true:
327 // CHECK2-NEXT:    br label [[COND_END:%.*]]
328 // CHECK2:       cond.false:
329 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
330 // CHECK2-NEXT:    br label [[COND_END]]
331 // CHECK2:       cond.end:
332 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
333 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
334 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
335 // CHECK2-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
336 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
337 // CHECK2:       omp.inner.for.cond:
338 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
339 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
340 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
341 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
342 // CHECK2:       omp.inner.for.body:
343 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
344 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
345 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
346 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
347 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[I]], align 4
348 // CHECK2-NEXT:    [[CMP3:%.*]] = icmp slt i32 [[TMP11]], 5
349 // CHECK2-NEXT:    br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
350 // CHECK2:       if.then:
351 // CHECK2-NEXT:    store i32 0, i32* [[TMP2]], align 4
352 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
353 // CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
354 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
355 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp sle i32 [[TMP13]], [[TMP12]]
356 // CHECK2-NEXT:    br i1 [[TMP14]], label [[LP_COND_THEN:%.*]], label [[LP_COND_EXIT:%.*]]
357 // CHECK2:       lp_cond_then:
358 // CHECK2-NEXT:    store i32 [[TMP12]], i32* @.{{pl_cond[.].+[.|,]}} align 4
359 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[TMP2]], align 4
360 // CHECK2-NEXT:    store i32 [[TMP15]], i32* @{{pl_cond[.].+[.|,]}} align 4
361 // CHECK2-NEXT:    br label [[LP_COND_EXIT]]
362 // CHECK2:       lp_cond_exit:
363 // CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
364 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 10)
365 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[TMP2]], i32* [[I]])
366 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
367 // CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
368 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
369 // CHECK2-NEXT:    [[TMP18:%.*]] = icmp sle i32 [[TMP17]], [[TMP16]]
370 // CHECK2-NEXT:    br i1 [[TMP18]], label [[LP_COND_THEN4:%.*]], label [[LP_COND_EXIT5:%.*]]
371 // CHECK2:       lp_cond_then4:
372 // CHECK2-NEXT:    store i32 [[TMP16]], i32* @.{{pl_cond[.].+[.|,]}} align 4
373 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP2]], align 4
374 // CHECK2-NEXT:    store i32 [[TMP19]], i32* @{{pl_cond[.].+[.|,]}} align 4
375 // CHECK2-NEXT:    br label [[LP_COND_EXIT5]]
376 // CHECK2:       lp_cond_exit5:
377 // CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
378 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
379 // CHECK2-NEXT:    [[TMP21:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP20]] monotonic, align 4
380 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
381 // CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
382 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
383 // CHECK2-NEXT:    [[TMP24:%.*]] = icmp sle i32 [[TMP23]], [[TMP22]]
384 // CHECK2-NEXT:    br i1 [[TMP24]], label [[LP_COND_THEN6:%.*]], label [[LP_COND_EXIT7:%.*]]
385 // CHECK2:       lp_cond_then6:
386 // CHECK2-NEXT:    store i32 [[TMP22]], i32* @.{{pl_cond[.].+[.|,]}} align 4
387 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[TMP2]], align 4
388 // CHECK2-NEXT:    store i32 [[TMP25]], i32* @{{pl_cond[.].+[.|,]}} align 4
389 // CHECK2-NEXT:    br label [[LP_COND_EXIT7]]
390 // CHECK2:       lp_cond_exit7:
391 // CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
392 // CHECK2-NEXT:    call void @__kmpc_push_num_threads(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 10)
393 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[TMP2]], i32* [[I]])
394 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL]], %struct.lasprivate.conditional* [[A1]], i32 0, i32 1
395 // CHECK2-NEXT:    [[TMP27:%.*]] = load i8, i8* [[TMP26]], align 4
396 // CHECK2-NEXT:    [[TMP28:%.*]] = icmp ne i8 [[TMP27]], 0
397 // CHECK2-NEXT:    br i1 [[TMP28]], label [[LPC_THEN:%.*]], label [[LPC_DONE:%.*]]
398 // CHECK2:       lpc.then:
399 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
400 // CHECK2-NEXT:    call void @__kmpc_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
401 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* @.{{pl_cond[.].+[.|,]}} align 4
402 // CHECK2-NEXT:    [[TMP31:%.*]] = icmp sle i32 [[TMP30]], [[TMP29]]
403 // CHECK2-NEXT:    br i1 [[TMP31]], label [[LP_COND_THEN8:%.*]], label [[LP_COND_EXIT9:%.*]]
404 // CHECK2:       lp_cond_then8:
405 // CHECK2-NEXT:    store i32 [[TMP29]], i32* @.{{pl_cond[.].+[.|,]}} align 4
406 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[TMP2]], align 4
407 // CHECK2-NEXT:    store i32 [[TMP32]], i32* @{{pl_cond[.].+[.|,]}} align 4
408 // CHECK2-NEXT:    br label [[LP_COND_EXIT9]]
409 // CHECK2:       lp_cond_exit9:
410 // CHECK2-NEXT:    call void @__kmpc_end_critical(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], [8 x i32]* @.gomp_critical_user_{{pl_cond[.].+[.|,]}}var)
411 // CHECK2-NEXT:    br label [[LPC_DONE]]
412 // CHECK2:       lpc.done:
413 // CHECK2-NEXT:    br label [[IF_END]]
414 // CHECK2:       if.end:
415 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
416 // CHECK2:       omp.body.continue:
417 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
418 // CHECK2:       omp.inner.for.inc:
419 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
420 // CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP33]], 1
421 // CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
422 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
423 // CHECK2:       omp.inner.for.end:
424 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
425 // CHECK2:       omp.loop.exit:
426 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
427 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
428 // CHECK2-NEXT:    [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
429 // CHECK2-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP4]])
430 // CHECK2-NEXT:    br i1 [[TMP35]], label [[DOTOMP_LASTPRIVATE_THEN:%.*]], label [[DOTOMP_LASTPRIVATE_DONE:%.*]]
431 // CHECK2:       .omp.lastprivate.then:
432 // CHECK2-NEXT:    [[TMP36:%.*]] = load i32, i32* @{{pl_cond[.].+[.|,]}} align 4
433 // CHECK2-NEXT:    store i32 [[TMP36]], i32* [[TMP2]], align 4
434 // CHECK2-NEXT:    [[TMP37:%.*]] = load i32, i32* [[TMP2]], align 4
435 // CHECK2-NEXT:    store i32 [[TMP37]], i32* [[TMP0]], align 4
436 // CHECK2-NEXT:    br label [[DOTOMP_LASTPRIVATE_DONE]]
437 // CHECK2:       .omp.lastprivate.done:
438 // CHECK2-NEXT:    ret void
439 //
440 //
441 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
442 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32* nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
443 // CHECK2-NEXT:  entry:
444 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
445 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
446 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
447 // CHECK2-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
448 // CHECK2-NEXT:    [[A1:%.*]] = alloca i32, align 4
449 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x i8*], align 8
450 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
451 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
452 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
453 // CHECK2-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
454 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
455 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[I_ADDR]], align 8
456 // CHECK2-NEXT:    store i32 0, i32* [[A1]], align 4
457 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
458 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[A1]], align 4
459 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP3]], [[TMP2]]
460 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[A1]], align 4
461 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
462 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i32* [[A1]] to i8*
463 // CHECK2-NEXT:    store i8* [[TMP5]], i8** [[TMP4]], align 8
464 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
465 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
466 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast [1 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
467 // CHECK2-NEXT:    [[TMP9:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP7]], i32 1, i64 8, i8* [[TMP8]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
468 // CHECK2-NEXT:    switch i32 [[TMP9]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
469 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
470 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
471 // CHECK2-NEXT:    ]
472 // CHECK2:       .omp.reduction.case1:
473 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP0]], align 4
474 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[A1]], align 4
475 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
476 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[TMP0]], align 4
477 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], [8 x i32]* @.gomp_critical_user_.reduction.var)
478 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
479 // CHECK2:       .omp.reduction.case2:
480 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[A1]], align 4
481 // CHECK2-NEXT:    [[TMP13:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP12]] monotonic, align 4
482 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
483 // CHECK2:       .omp.reduction.default:
484 // CHECK2-NEXT:    ret void
485 //
486 //
487 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
488 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
489 // CHECK2-NEXT:  entry:
490 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
491 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
492 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
493 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
494 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
495 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [1 x i8*]*
496 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
497 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [1 x i8*]*
498 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP5]], i64 0, i64 0
499 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
500 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
501 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[TMP3]], i64 0, i64 0
502 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
503 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
504 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
505 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP8]], align 4
506 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
507 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
508 // CHECK2-NEXT:    ret void
509 //
510 //
511 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
512 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32* nonnull align 4 dereferenceable(4) [[I:%.*]]) #[[ATTR1]] {
513 // CHECK2-NEXT:  entry:
514 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
515 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
516 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca i32*, align 8
517 // CHECK2-NEXT:    [[I_ADDR:%.*]] = alloca i32*, align 8
518 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
519 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
520 // CHECK2-NEXT:    store i32* [[A]], i32** [[A_ADDR]], align 8
521 // CHECK2-NEXT:    store i32* [[I]], i32** [[I_ADDR]], align 8
522 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[A_ADDR]], align 8
523 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[I_ADDR]], align 8
524 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
525 // CHECK2-NEXT:    [[TMP3:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP2]] monotonic, align 4
526 // CHECK2-NEXT:    [[TMP4:%.*]] = bitcast i32* [[TMP0]] to %struct.lasprivate.conditional*
527 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_LASPRIVATE_CONDITIONAL:%.*]], %struct.lasprivate.conditional* [[TMP4]], i32 0, i32 1
528 // CHECK2-NEXT:    store atomic volatile i8 1, i8* [[TMP5]] unordered, align 1
529 // CHECK2-NEXT:    ret void
530 //
531 //