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 -fopenmp-version=50 -x c++ -triple x86_64-unknown-linux -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5 
6 // RUN: %clang_cc1 -triple x86_64-unknown-linux -verify -fopenmp-simd -fopenmp-version=50 -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -debug-info-kind=limited -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // expected-no-diagnostics
10 #ifndef HEADER
11 #define HEADER
12 
13 int main(int argc, char **argv) {
14 #pragma omp parallel reduction(task, +: argc, argv[0:10][0:argc])
15   {
16 #pragma omp task in_reduction(+: argc, argv[0:10][0:argc])
17     ;
18   }
19 }
20 
21 
22 
23 // Init firstprivate copy of argc
24 
25 // Init firstprivate copy of argv[0:10][0:argc]
26 
27 // Register task reduction.
28 
29 
30 
31 
32 
33 
34 
35 
36 
37 #endif
38 // CHECK1-LABEL: define {{[^@]+}}@main
39 // CHECK1-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
40 // CHECK1-NEXT:  entry:
41 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
42 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
43 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
44 // CHECK1-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
45 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
46 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
47 // CHECK1-NEXT:    store i32* [[ARGC_ADDR]], i32** [[TMP0]], align 8
48 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
49 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
50 // CHECK1-NEXT:    store i8** [[TMP2]], i8*** [[TMP1]], align 8
51 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[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_]])
52 // CHECK1-NEXT:    ret i32 0
53 //
54 //
55 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
56 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
57 // CHECK1-NEXT:  entry:
58 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
59 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
60 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
61 // CHECK1-NEXT:    [[ARGC:%.*]] = alloca i32, align 4
62 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
63 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
64 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
65 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i8*, align 8
66 // CHECK1-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
67 // CHECK1-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
68 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
69 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
70 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
71 // CHECK1-NEXT:    [[_TMP23:%.*]] = alloca i8, align 1
72 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
73 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
74 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
75 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
76 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
77 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
78 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
79 // CHECK1-NEXT:    store i32 0, i32* [[ARGC]], align 4
80 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8**, i8*** [[TMP3]], align 8
81 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP4]], i64 0
82 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
83 // CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i64 0
84 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP2]], align 4
85 // CHECK1-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
86 // CHECK1-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
87 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP3]], align 8
88 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 9
89 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX2]], align 8
90 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 [[LB_ADD_LEN]]
91 // CHECK1-NEXT:    [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX3]] to i64
92 // CHECK1-NEXT:    [[TMP11:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
93 // CHECK1-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
94 // CHECK1-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
95 // CHECK1-NEXT:    [[TMP14:%.*]] = add nuw i64 [[TMP13]], 1
96 // CHECK1-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
97 // CHECK1-NEXT:    [[TMP16:%.*]] = call i8* @llvm.stacksave()
98 // CHECK1-NEXT:    store i8* [[TMP16]], i8** [[SAVED_STACK]], align 8
99 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP14]], align 16
100 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[__VLA_EXPR0]], align 8
101 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP14]]
102 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP17]]
103 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
104 // CHECK1:       omp.arrayinit.body:
105 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
106 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
107 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
108 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
109 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
110 // CHECK1:       omp.arrayinit.done:
111 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP3]], align 8
112 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
113 // CHECK1-NEXT:    [[TMP20:%.*]] = ptrtoint i8* [[TMP19]] to i64
114 // CHECK1-NEXT:    [[TMP21:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
115 // CHECK1-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
116 // CHECK1-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
117 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP23]]
118 // CHECK1-NEXT:    store i8** [[_TMP4]], i8*** [[TMP]], align 8
119 // CHECK1-NEXT:    store i8* [[TMP24]], i8** [[_TMP4]], align 8
120 // CHECK1-NEXT:    [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 0
121 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
122 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i32* [[ARGC]] to i8*
123 // CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
124 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
125 // CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP2]] to i8*
126 // CHECK1-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
127 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
128 // CHECK1-NEXT:    store i64 4, i64* [[TMP29]], align 8
129 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
130 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP30]], align 8
131 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
132 // CHECK1-NEXT:    store i8* null, i8** [[TMP31]], align 8
133 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
134 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP32]], align 8
135 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
136 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i32* [[TMP33]] to i8*
137 // CHECK1-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP34]], i8 0, i64 4, i1 false)
138 // CHECK1-NEXT:    [[DOTRD_INPUT_GEP_5:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
139 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 0
140 // CHECK1-NEXT:    [[TMP36:%.*]] = load i8**, i8*** [[TMP3]], align 8
141 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP36]], i64 0
142 // CHECK1-NEXT:    [[TMP37:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
143 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP37]], i64 0
144 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP2]], align 4
145 // CHECK1-NEXT:    [[TMP39:%.*]] = sext i32 [[TMP38]] to i64
146 // CHECK1-NEXT:    [[LB_ADD_LEN8:%.*]] = add nsw i64 -1, [[TMP39]]
147 // CHECK1-NEXT:    [[TMP40:%.*]] = load i8**, i8*** [[TMP3]], align 8
148 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8*, i8** [[TMP40]], i64 9
149 // CHECK1-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[ARRAYIDX9]], align 8
150 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, i8* [[TMP41]], i64 [[LB_ADD_LEN8]]
151 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP35]], align 8
152 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 1
153 // CHECK1-NEXT:    store i8* [[ARRAYIDX7]], i8** [[TMP42]], align 8
154 // CHECK1-NEXT:    [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX10]] to i64
155 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[ARRAYIDX7]] to i64
156 // CHECK1-NEXT:    [[TMP45:%.*]] = sub i64 [[TMP43]], [[TMP44]]
157 // CHECK1-NEXT:    [[TMP46:%.*]] = sdiv exact i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
158 // CHECK1-NEXT:    [[TMP47:%.*]] = add nuw i64 [[TMP46]], 1
159 // CHECK1-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
160 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 2
161 // CHECK1-NEXT:    store i64 [[TMP48]], i64* [[TMP49]], align 8
162 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 3
163 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP50]], align 8
164 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 4
165 // CHECK1-NEXT:    store i8* null, i8** [[TMP51]], align 8
166 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 5
167 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP52]], align 8
168 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 6
169 // CHECK1-NEXT:    store i32 1, i32* [[TMP53]], align 8
170 // CHECK1-NEXT:    [[TMP54:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
171 // CHECK1-NEXT:    [[TMP55:%.*]] = load i32, i32* [[TMP54]], align 4
172 // CHECK1-NEXT:    [[TMP56:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
173 // CHECK1-NEXT:    [[TMP57:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP55]], i32 0, i32 2, i8* [[TMP56]])
174 // CHECK1-NEXT:    store i8* [[TMP57]], i8** [[DOTTASK_RED_]], align 8
175 // CHECK1-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
176 // CHECK1-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP58]], align 8
177 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
178 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[TMP59]], align 8
179 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
180 // CHECK1-NEXT:    [[TMP61:%.*]] = load i8**, i8*** [[TMP]], align 8
181 // CHECK1-NEXT:    store i8** [[TMP61]], i8*** [[TMP60]], align 8
182 // CHECK1-NEXT:    [[TMP62:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
183 // CHECK1-NEXT:    [[TMP63:%.*]] = load i32, i32* [[TMP62]], align 4
184 // CHECK1-NEXT:    [[TMP64:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP63]], i32 1, i64 48, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
185 // CHECK1-NEXT:    [[TMP65:%.*]] = bitcast i8* [[TMP64]] to %struct.kmp_task_t_with_privates*
186 // CHECK1-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP65]], i32 0, i32 0
187 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP66]], i32 0, i32 0
188 // CHECK1-NEXT:    [[TMP68:%.*]] = load i8*, i8** [[TMP67]], align 8
189 // CHECK1-NEXT:    [[TMP69:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
190 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP68]], i8* align 8 [[TMP69]], i64 24, i1 false)
191 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP65]], i32 0, i32 1
192 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP70]], i32 0, i32 0
193 // CHECK1-NEXT:    [[TMP72:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
194 // CHECK1-NEXT:    store i8* [[TMP72]], i8** [[TMP71]], align 8
195 // CHECK1-NEXT:    [[TMP73:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
196 // CHECK1-NEXT:    [[TMP74:%.*]] = load i32, i32* [[TMP73]], align 4
197 // CHECK1-NEXT:    [[TMP75:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP74]], i8* [[TMP64]])
198 // CHECK1-NEXT:    [[TMP76:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
199 // CHECK1-NEXT:    [[TMP77:%.*]] = load i32, i32* [[TMP76]], align 4
200 // CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP77]], i32 0)
201 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
202 // CHECK1-NEXT:    [[TMP79:%.*]] = bitcast i32* [[ARGC]] to i8*
203 // CHECK1-NEXT:    store i8* [[TMP79]], i8** [[TMP78]], align 8
204 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
205 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP80]], align 8
206 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
207 // CHECK1-NEXT:    [[TMP82:%.*]] = inttoptr i64 [[TMP14]] to i8*
208 // CHECK1-NEXT:    store i8* [[TMP82]], i8** [[TMP81]], align 8
209 // CHECK1-NEXT:    [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
210 // CHECK1-NEXT:    [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
211 // CHECK1-NEXT:    [[TMP85:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
212 // CHECK1-NEXT:    [[TMP86:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP84]], i32 2, i64 24, i8* [[TMP85]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
213 // CHECK1-NEXT:    switch i32 [[TMP86]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
214 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
215 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
216 // CHECK1-NEXT:    ]
217 // CHECK1:       .omp.reduction.case1:
218 // CHECK1-NEXT:    [[TMP87:%.*]] = load i32, i32* [[TMP2]], align 4
219 // CHECK1-NEXT:    [[TMP88:%.*]] = load i32, i32* [[ARGC]], align 4
220 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP87]], [[TMP88]]
221 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
222 // CHECK1-NEXT:    [[TMP89:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
223 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP89]]
224 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
225 // CHECK1:       omp.arraycpy.body:
226 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
227 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
228 // CHECK1-NEXT:    [[TMP90:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
229 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP90]] to i32
230 // CHECK1-NEXT:    [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
231 // CHECK1-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP91]] to i32
232 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
233 // CHECK1-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
234 // CHECK1-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
235 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
236 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
237 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP89]]
238 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
239 // CHECK1:       omp.arraycpy.done17:
240 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]], [8 x i32]* @.gomp_critical_user_.reduction.var)
241 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
242 // CHECK1:       .omp.reduction.case2:
243 // CHECK1-NEXT:    [[TMP92:%.*]] = load i32, i32* [[ARGC]], align 4
244 // CHECK1-NEXT:    [[TMP93:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP92]] monotonic, align 4
245 // CHECK1-NEXT:    [[TMP94:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
246 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP94]]
247 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
248 // CHECK1:       omp.arraycpy.body19:
249 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
250 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
251 // CHECK1-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
252 // CHECK1-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP95]] to i32
253 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
254 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
255 // CHECK1:       atomic_cont:
256 // CHECK1-NEXT:    [[TMP96:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP101:%.*]], [[ATOMIC_CONT]] ]
257 // CHECK1-NEXT:    store i8 [[TMP96]], i8* [[_TMP23]], align 1
258 // CHECK1-NEXT:    [[TMP97:%.*]] = load i8, i8* [[_TMP23]], align 1
259 // CHECK1-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP97]] to i32
260 // CHECK1-NEXT:    [[TMP98:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
261 // CHECK1-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP98]] to i32
262 // CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
263 // CHECK1-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
264 // CHECK1-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
265 // CHECK1-NEXT:    [[TMP99:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
266 // CHECK1-NEXT:    [[TMP100:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP96]], i8 [[TMP99]] monotonic monotonic, align 1
267 // CHECK1-NEXT:    [[TMP101]] = extractvalue { i8, i1 } [[TMP100]], 0
268 // CHECK1-NEXT:    [[TMP102:%.*]] = extractvalue { i8, i1 } [[TMP100]], 1
269 // CHECK1-NEXT:    br i1 [[TMP102]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
270 // CHECK1:       atomic_exit:
271 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
272 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
273 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP94]]
274 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
275 // CHECK1:       omp.arraycpy.done31:
276 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
277 // CHECK1:       .omp.reduction.default:
278 // CHECK1-NEXT:    [[TMP103:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
279 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP103]])
280 // CHECK1-NEXT:    ret void
281 //
282 //
283 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
284 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
285 // CHECK1-NEXT:  entry:
286 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
287 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
288 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
289 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
290 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
291 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
292 // CHECK1-NEXT:    store i32 0, i32* [[TMP3]], align 8
293 // CHECK1-NEXT:    ret void
294 //
295 //
296 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
297 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
298 // CHECK1-NEXT:  entry:
299 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
300 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
301 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
302 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
303 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
304 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
305 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
306 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
307 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
308 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
309 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
310 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
311 // CHECK1-NEXT:    ret void
312 //
313 //
314 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
315 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
316 // CHECK1-NEXT:  entry:
317 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
318 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
319 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
320 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
321 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
322 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
323 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
324 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
325 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
326 // CHECK1:       omp.arrayinit.body:
327 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
328 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
329 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
330 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
331 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
332 // CHECK1:       omp.arrayinit.done:
333 // CHECK1-NEXT:    ret void
334 //
335 //
336 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
337 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
338 // CHECK1-NEXT:  entry:
339 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
340 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
341 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
342 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
343 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
344 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
345 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
346 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
347 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
348 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
349 // CHECK1:       omp.arraycpy.body:
350 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
351 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
352 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
353 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
354 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
355 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
356 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
357 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
358 // CHECK1-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
359 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
360 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
361 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
362 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
363 // CHECK1:       omp.arraycpy.done4:
364 // CHECK1-NEXT:    ret void
365 //
366 //
367 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
368 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
369 // CHECK1-NEXT:  entry:
370 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
371 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
372 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
373 // CHECK1-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
374 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
375 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
376 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
377 // CHECK1-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
378 // CHECK1-NEXT:    ret void
379 //
380 //
381 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
382 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
383 // CHECK1-NEXT:  entry:
384 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
386 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
387 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
388 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
389 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
390 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
391 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
392 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
393 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
394 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
395 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
396 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
397 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
398 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
399 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
400 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
401 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
402 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
403 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
404 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
405 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
406 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
407 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
408 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
409 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
410 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
411 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
412 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
413 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
414 // CHECK1-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i8***)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
415 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
416 // CHECK1-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
417 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
418 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
419 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
420 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
421 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
422 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
423 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
424 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
425 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
426 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
427 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
428 // CHECK1-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
429 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
430 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
431 // CHECK1-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
432 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
433 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
434 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
435 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
436 // CHECK1-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
437 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
438 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
439 // CHECK1-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
440 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
441 // CHECK1-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
442 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
443 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
444 // CHECK1-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
445 // CHECK1-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
446 // CHECK1-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
447 // CHECK1-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
448 // CHECK1-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
449 // CHECK1-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
450 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
451 // CHECK1-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
452 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
453 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
454 // CHECK1-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
455 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
456 // CHECK1-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
457 // CHECK1-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
458 // CHECK1-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
459 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
460 // CHECK1-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
461 // CHECK1-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
462 // CHECK1-NEXT:    ret i32 0
463 //
464 //
465 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
466 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
467 // CHECK1-NEXT:  entry:
468 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
469 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
470 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
471 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
472 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
473 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
474 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
475 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
476 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
477 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
478 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
479 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
480 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
481 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
482 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
483 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
484 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
485 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
486 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
487 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
488 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
489 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
490 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
491 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
492 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
493 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
494 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
495 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
496 // CHECK1:       omp.arraycpy.body:
497 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
498 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
499 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
500 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
501 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
502 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
503 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
504 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
505 // CHECK1-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
506 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
507 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
508 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
509 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
510 // CHECK1:       omp.arraycpy.done5:
511 // CHECK1-NEXT:    ret void
512 //
513 //
514 // CHECK2-LABEL: define {{[^@]+}}@main
515 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
516 // CHECK2-NEXT:  entry:
517 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
518 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
519 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
520 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
521 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
522 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
523 // CHECK2-NEXT:    store i32* [[ARGC_ADDR]], i32** [[TMP0]], align 8
524 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
525 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
526 // CHECK2-NEXT:    store i8** [[TMP2]], i8*** [[TMP1]], align 8
527 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[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_]])
528 // CHECK2-NEXT:    ret i32 0
529 //
530 //
531 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
532 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
533 // CHECK2-NEXT:  entry:
534 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
535 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
536 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
537 // CHECK2-NEXT:    [[ARGC:%.*]] = alloca i32, align 4
538 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
539 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
540 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
541 // CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i8*, align 8
542 // CHECK2-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
543 // CHECK2-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
544 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
545 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
546 // CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
547 // CHECK2-NEXT:    [[_TMP23:%.*]] = alloca i8, align 1
548 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
549 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
550 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
551 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
552 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
553 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
554 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
555 // CHECK2-NEXT:    store i32 0, i32* [[ARGC]], align 4
556 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8**, i8*** [[TMP3]], align 8
557 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP4]], i64 0
558 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
559 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i64 0
560 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP2]], align 4
561 // CHECK2-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
562 // CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
563 // CHECK2-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP3]], align 8
564 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 9
565 // CHECK2-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX2]], align 8
566 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 [[LB_ADD_LEN]]
567 // CHECK2-NEXT:    [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX3]] to i64
568 // CHECK2-NEXT:    [[TMP11:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
569 // CHECK2-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
570 // CHECK2-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
571 // CHECK2-NEXT:    [[TMP14:%.*]] = add nuw i64 [[TMP13]], 1
572 // CHECK2-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
573 // CHECK2-NEXT:    [[TMP16:%.*]] = call i8* @llvm.stacksave()
574 // CHECK2-NEXT:    store i8* [[TMP16]], i8** [[SAVED_STACK]], align 8
575 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP14]], align 16
576 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[__VLA_EXPR0]], align 8
577 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP14]]
578 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP17]]
579 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
580 // CHECK2:       omp.arrayinit.body:
581 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
582 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
583 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
584 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
585 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
586 // CHECK2:       omp.arrayinit.done:
587 // CHECK2-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP3]], align 8
588 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
589 // CHECK2-NEXT:    [[TMP20:%.*]] = ptrtoint i8* [[TMP19]] to i64
590 // CHECK2-NEXT:    [[TMP21:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
591 // CHECK2-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
592 // CHECK2-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
593 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP23]]
594 // CHECK2-NEXT:    store i8** [[_TMP4]], i8*** [[TMP]], align 8
595 // CHECK2-NEXT:    store i8* [[TMP24]], i8** [[_TMP4]], align 8
596 // CHECK2-NEXT:    [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 0
597 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
598 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[ARGC]] to i8*
599 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
600 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
601 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP2]] to i8*
602 // CHECK2-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
603 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
604 // CHECK2-NEXT:    store i64 4, i64* [[TMP29]], align 8
605 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
606 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP30]], align 8
607 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
608 // CHECK2-NEXT:    store i8* null, i8** [[TMP31]], align 8
609 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
610 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP32]], align 8
611 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
612 // CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i32* [[TMP33]] to i8*
613 // CHECK2-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP34]], i8 0, i64 4, i1 false)
614 // CHECK2-NEXT:    [[DOTRD_INPUT_GEP_5:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
615 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 0
616 // CHECK2-NEXT:    [[TMP36:%.*]] = load i8**, i8*** [[TMP3]], align 8
617 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP36]], i64 0
618 // CHECK2-NEXT:    [[TMP37:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
619 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP37]], i64 0
620 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP2]], align 4
621 // CHECK2-NEXT:    [[TMP39:%.*]] = sext i32 [[TMP38]] to i64
622 // CHECK2-NEXT:    [[LB_ADD_LEN8:%.*]] = add nsw i64 -1, [[TMP39]]
623 // CHECK2-NEXT:    [[TMP40:%.*]] = load i8**, i8*** [[TMP3]], align 8
624 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8*, i8** [[TMP40]], i64 9
625 // CHECK2-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[ARRAYIDX9]], align 8
626 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, i8* [[TMP41]], i64 [[LB_ADD_LEN8]]
627 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP35]], align 8
628 // CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 1
629 // CHECK2-NEXT:    store i8* [[ARRAYIDX7]], i8** [[TMP42]], align 8
630 // CHECK2-NEXT:    [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX10]] to i64
631 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[ARRAYIDX7]] to i64
632 // CHECK2-NEXT:    [[TMP45:%.*]] = sub i64 [[TMP43]], [[TMP44]]
633 // CHECK2-NEXT:    [[TMP46:%.*]] = sdiv exact i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
634 // CHECK2-NEXT:    [[TMP47:%.*]] = add nuw i64 [[TMP46]], 1
635 // CHECK2-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
636 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 2
637 // CHECK2-NEXT:    store i64 [[TMP48]], i64* [[TMP49]], align 8
638 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 3
639 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP50]], align 8
640 // CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 4
641 // CHECK2-NEXT:    store i8* null, i8** [[TMP51]], align 8
642 // CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 5
643 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP52]], align 8
644 // CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 6
645 // CHECK2-NEXT:    store i32 1, i32* [[TMP53]], align 8
646 // CHECK2-NEXT:    [[TMP54:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
647 // CHECK2-NEXT:    [[TMP55:%.*]] = load i32, i32* [[TMP54]], align 4
648 // CHECK2-NEXT:    [[TMP56:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
649 // CHECK2-NEXT:    [[TMP57:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP55]], i32 0, i32 2, i8* [[TMP56]])
650 // CHECK2-NEXT:    store i8* [[TMP57]], i8** [[DOTTASK_RED_]], align 8
651 // CHECK2-NEXT:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
652 // CHECK2-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP58]], align 8
653 // CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
654 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[TMP59]], align 8
655 // CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
656 // CHECK2-NEXT:    [[TMP61:%.*]] = load i8**, i8*** [[TMP]], align 8
657 // CHECK2-NEXT:    store i8** [[TMP61]], i8*** [[TMP60]], align 8
658 // CHECK2-NEXT:    [[TMP62:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
659 // CHECK2-NEXT:    [[TMP63:%.*]] = load i32, i32* [[TMP62]], align 4
660 // CHECK2-NEXT:    [[TMP64:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP63]], i32 1, i64 48, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
661 // CHECK2-NEXT:    [[TMP65:%.*]] = bitcast i8* [[TMP64]] to %struct.kmp_task_t_with_privates*
662 // CHECK2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP65]], i32 0, i32 0
663 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP66]], i32 0, i32 0
664 // CHECK2-NEXT:    [[TMP68:%.*]] = load i8*, i8** [[TMP67]], align 8
665 // CHECK2-NEXT:    [[TMP69:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
666 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP68]], i8* align 8 [[TMP69]], i64 24, i1 false)
667 // CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP65]], i32 0, i32 1
668 // CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP70]], i32 0, i32 0
669 // CHECK2-NEXT:    [[TMP72:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
670 // CHECK2-NEXT:    store i8* [[TMP72]], i8** [[TMP71]], align 8
671 // CHECK2-NEXT:    [[TMP73:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
672 // CHECK2-NEXT:    [[TMP74:%.*]] = load i32, i32* [[TMP73]], align 4
673 // CHECK2-NEXT:    [[TMP75:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP74]], i8* [[TMP64]])
674 // CHECK2-NEXT:    [[TMP76:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
675 // CHECK2-NEXT:    [[TMP77:%.*]] = load i32, i32* [[TMP76]], align 4
676 // CHECK2-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP77]], i32 0)
677 // CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
678 // CHECK2-NEXT:    [[TMP79:%.*]] = bitcast i32* [[ARGC]] to i8*
679 // CHECK2-NEXT:    store i8* [[TMP79]], i8** [[TMP78]], align 8
680 // CHECK2-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
681 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP80]], align 8
682 // CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
683 // CHECK2-NEXT:    [[TMP82:%.*]] = inttoptr i64 [[TMP14]] to i8*
684 // CHECK2-NEXT:    store i8* [[TMP82]], i8** [[TMP81]], align 8
685 // CHECK2-NEXT:    [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
686 // CHECK2-NEXT:    [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
687 // CHECK2-NEXT:    [[TMP85:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
688 // CHECK2-NEXT:    [[TMP86:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP84]], i32 2, i64 24, i8* [[TMP85]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
689 // CHECK2-NEXT:    switch i32 [[TMP86]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
690 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
691 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
692 // CHECK2-NEXT:    ]
693 // CHECK2:       .omp.reduction.case1:
694 // CHECK2-NEXT:    [[TMP87:%.*]] = load i32, i32* [[TMP2]], align 4
695 // CHECK2-NEXT:    [[TMP88:%.*]] = load i32, i32* [[ARGC]], align 4
696 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP87]], [[TMP88]]
697 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
698 // CHECK2-NEXT:    [[TMP89:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
699 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP89]]
700 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
701 // CHECK2:       omp.arraycpy.body:
702 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
703 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
704 // CHECK2-NEXT:    [[TMP90:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
705 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP90]] to i32
706 // CHECK2-NEXT:    [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
707 // CHECK2-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP91]] to i32
708 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
709 // CHECK2-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
710 // CHECK2-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
711 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
712 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
713 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP89]]
714 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
715 // CHECK2:       omp.arraycpy.done17:
716 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]], [8 x i32]* @.gomp_critical_user_.reduction.var)
717 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
718 // CHECK2:       .omp.reduction.case2:
719 // CHECK2-NEXT:    [[TMP92:%.*]] = load i32, i32* [[ARGC]], align 4
720 // CHECK2-NEXT:    [[TMP93:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP92]] monotonic, align 4
721 // CHECK2-NEXT:    [[TMP94:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
722 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP94]]
723 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
724 // CHECK2:       omp.arraycpy.body19:
725 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
726 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
727 // CHECK2-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
728 // CHECK2-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP95]] to i32
729 // CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
730 // CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
731 // CHECK2:       atomic_cont:
732 // CHECK2-NEXT:    [[TMP96:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP101:%.*]], [[ATOMIC_CONT]] ]
733 // CHECK2-NEXT:    store i8 [[TMP96]], i8* [[_TMP23]], align 1
734 // CHECK2-NEXT:    [[TMP97:%.*]] = load i8, i8* [[_TMP23]], align 1
735 // CHECK2-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP97]] to i32
736 // CHECK2-NEXT:    [[TMP98:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
737 // CHECK2-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP98]] to i32
738 // CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
739 // CHECK2-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
740 // CHECK2-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
741 // CHECK2-NEXT:    [[TMP99:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
742 // CHECK2-NEXT:    [[TMP100:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP96]], i8 [[TMP99]] monotonic monotonic, align 1
743 // CHECK2-NEXT:    [[TMP101]] = extractvalue { i8, i1 } [[TMP100]], 0
744 // CHECK2-NEXT:    [[TMP102:%.*]] = extractvalue { i8, i1 } [[TMP100]], 1
745 // CHECK2-NEXT:    br i1 [[TMP102]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
746 // CHECK2:       atomic_exit:
747 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
748 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
749 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP94]]
750 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
751 // CHECK2:       omp.arraycpy.done31:
752 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
753 // CHECK2:       .omp.reduction.default:
754 // CHECK2-NEXT:    [[TMP103:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
755 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP103]])
756 // CHECK2-NEXT:    ret void
757 //
758 //
759 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
760 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
761 // CHECK2-NEXT:  entry:
762 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
763 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
764 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
765 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
766 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
767 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
768 // CHECK2-NEXT:    store i32 0, i32* [[TMP3]], align 8
769 // CHECK2-NEXT:    ret void
770 //
771 //
772 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
773 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
774 // CHECK2-NEXT:  entry:
775 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
776 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
777 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
778 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
779 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
780 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
781 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
782 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
783 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
784 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
785 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
786 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
787 // CHECK2-NEXT:    ret void
788 //
789 //
790 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
791 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
792 // CHECK2-NEXT:  entry:
793 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
794 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
795 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
796 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
797 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
798 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
799 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
800 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
801 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
802 // CHECK2:       omp.arrayinit.body:
803 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
804 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
805 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
806 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
807 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
808 // CHECK2:       omp.arrayinit.done:
809 // CHECK2-NEXT:    ret void
810 //
811 //
812 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
813 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
814 // CHECK2-NEXT:  entry:
815 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
816 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
817 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
818 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
819 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
820 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
821 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
822 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
823 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
824 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
825 // CHECK2:       omp.arraycpy.body:
826 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
827 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
828 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
829 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
830 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
831 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
832 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
833 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
834 // CHECK2-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
835 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
836 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
837 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
838 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
839 // CHECK2:       omp.arraycpy.done4:
840 // CHECK2-NEXT:    ret void
841 //
842 //
843 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
844 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
845 // CHECK2-NEXT:  entry:
846 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
847 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
848 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
849 // CHECK2-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
850 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
851 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
852 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
853 // CHECK2-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
854 // CHECK2-NEXT:    ret void
855 //
856 //
857 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
858 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
859 // CHECK2-NEXT:  entry:
860 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
861 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
862 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
863 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
864 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
865 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
866 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
867 // CHECK2-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
868 // CHECK2-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
869 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
870 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
871 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
872 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
873 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
874 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
875 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
876 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
877 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
878 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
879 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
880 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
881 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
882 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
883 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
884 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
885 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
886 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
887 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
888 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
889 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
890 // CHECK2-NEXT:    store void (i8*, ...)* bitcast (void (%struct..kmp_privates.t*, i8***)* @.omp_task_privates_map. to void (i8*, ...)*), void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
891 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
892 // CHECK2-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
893 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
894 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
895 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
896 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
897 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
898 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
899 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
900 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
901 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
902 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
903 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
904 // CHECK2-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
905 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
906 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
907 // CHECK2-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
908 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
909 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
910 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
911 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
912 // CHECK2-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
913 // CHECK2-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
914 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
915 // CHECK2-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
916 // CHECK2-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
917 // CHECK2-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
918 // CHECK2-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
919 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
920 // CHECK2-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
921 // CHECK2-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
922 // CHECK2-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
923 // CHECK2-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
924 // CHECK2-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
925 // CHECK2-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
926 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
927 // CHECK2-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
928 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
929 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
930 // CHECK2-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
931 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
932 // CHECK2-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
933 // CHECK2-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
934 // CHECK2-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
935 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
936 // CHECK2-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
937 // CHECK2-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
938 // CHECK2-NEXT:    ret i32 0
939 //
940 //
941 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
942 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
943 // CHECK2-NEXT:  entry:
944 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
945 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
946 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
947 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
948 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
949 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
950 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
951 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
952 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
953 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
954 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
955 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
956 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
957 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
958 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
959 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
960 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
961 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
962 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
963 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
964 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
965 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
966 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
967 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
968 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
969 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
970 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
971 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
972 // CHECK2:       omp.arraycpy.body:
973 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
974 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
975 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
976 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
977 // CHECK2-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
978 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
979 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
980 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
981 // CHECK2-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
982 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
983 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
984 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
985 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
986 // CHECK2:       omp.arraycpy.done5:
987 // CHECK2-NEXT:    ret void
988 //
989