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 sections 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:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
62 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
63 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
64 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
65 // CHECK1-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
66 // CHECK1-NEXT:    [[ARGC:%.*]] = alloca i32, align 4
67 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
68 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
69 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
70 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i8*, align 8
71 // CHECK1-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
72 // CHECK1-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
73 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
74 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
75 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
76 // CHECK1-NEXT:    [[_TMP23:%.*]] = alloca i8, align 1
77 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
78 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
79 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
80 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
81 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
82 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
83 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
84 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
85 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
86 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
87 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
88 // CHECK1-NEXT:    store i32 0, i32* [[ARGC]], align 4
89 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8**, i8*** [[TMP3]], align 8
90 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP4]], i64 0
91 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
92 // CHECK1-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i64 0
93 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP2]], align 4
94 // CHECK1-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
95 // CHECK1-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
96 // CHECK1-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP3]], align 8
97 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 9
98 // CHECK1-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX2]], align 8
99 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 [[LB_ADD_LEN]]
100 // CHECK1-NEXT:    [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX3]] to i64
101 // CHECK1-NEXT:    [[TMP11:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
102 // CHECK1-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
103 // CHECK1-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
104 // CHECK1-NEXT:    [[TMP14:%.*]] = add nuw i64 [[TMP13]], 1
105 // CHECK1-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
106 // CHECK1-NEXT:    [[TMP16:%.*]] = call i8* @llvm.stacksave()
107 // CHECK1-NEXT:    store i8* [[TMP16]], i8** [[SAVED_STACK]], align 8
108 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP14]], align 16
109 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[__VLA_EXPR0]], align 8
110 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP14]]
111 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP17]]
112 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
113 // CHECK1:       omp.arrayinit.body:
114 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
115 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
116 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
117 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
118 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
119 // CHECK1:       omp.arrayinit.done:
120 // CHECK1-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP3]], align 8
121 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
122 // CHECK1-NEXT:    [[TMP20:%.*]] = ptrtoint i8* [[TMP19]] to i64
123 // CHECK1-NEXT:    [[TMP21:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
124 // CHECK1-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
125 // CHECK1-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
126 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP23]]
127 // CHECK1-NEXT:    store i8** [[_TMP4]], i8*** [[TMP]], align 8
128 // CHECK1-NEXT:    store i8* [[TMP24]], i8** [[_TMP4]], align 8
129 // 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
130 // CHECK1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
131 // CHECK1-NEXT:    [[TMP26:%.*]] = bitcast i32* [[ARGC]] to i8*
132 // CHECK1-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
133 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
134 // CHECK1-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP2]] to i8*
135 // CHECK1-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
136 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
137 // CHECK1-NEXT:    store i64 4, i64* [[TMP29]], align 8
138 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
139 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP30]], align 8
140 // CHECK1-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
141 // CHECK1-NEXT:    store i8* null, i8** [[TMP31]], align 8
142 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
143 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP32]], align 8
144 // CHECK1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
145 // CHECK1-NEXT:    [[TMP34:%.*]] = bitcast i32* [[TMP33]] to i8*
146 // CHECK1-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP34]], i8 0, i64 4, i1 false)
147 // 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
148 // CHECK1-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 0
149 // CHECK1-NEXT:    [[TMP36:%.*]] = load i8**, i8*** [[TMP3]], align 8
150 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP36]], i64 0
151 // CHECK1-NEXT:    [[TMP37:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
152 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP37]], i64 0
153 // CHECK1-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP2]], align 4
154 // CHECK1-NEXT:    [[TMP39:%.*]] = sext i32 [[TMP38]] to i64
155 // CHECK1-NEXT:    [[LB_ADD_LEN8:%.*]] = add nsw i64 -1, [[TMP39]]
156 // CHECK1-NEXT:    [[TMP40:%.*]] = load i8**, i8*** [[TMP3]], align 8
157 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8*, i8** [[TMP40]], i64 9
158 // CHECK1-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[ARRAYIDX9]], align 8
159 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, i8* [[TMP41]], i64 [[LB_ADD_LEN8]]
160 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP35]], align 8
161 // CHECK1-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 1
162 // CHECK1-NEXT:    store i8* [[ARRAYIDX7]], i8** [[TMP42]], align 8
163 // CHECK1-NEXT:    [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX10]] to i64
164 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[ARRAYIDX7]] to i64
165 // CHECK1-NEXT:    [[TMP45:%.*]] = sub i64 [[TMP43]], [[TMP44]]
166 // CHECK1-NEXT:    [[TMP46:%.*]] = sdiv exact i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
167 // CHECK1-NEXT:    [[TMP47:%.*]] = add nuw i64 [[TMP46]], 1
168 // CHECK1-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
169 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 2
170 // CHECK1-NEXT:    store i64 [[TMP48]], i64* [[TMP49]], align 8
171 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 3
172 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP50]], align 8
173 // CHECK1-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 4
174 // CHECK1-NEXT:    store i8* null, i8** [[TMP51]], align 8
175 // CHECK1-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 5
176 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP52]], align 8
177 // CHECK1-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 6
178 // CHECK1-NEXT:    store i32 1, i32* [[TMP53]], align 8
179 // CHECK1-NEXT:    [[TMP54:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
180 // CHECK1-NEXT:    [[TMP55:%.*]] = load i32, i32* [[TMP54]], align 4
181 // CHECK1-NEXT:    [[TMP56:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
182 // CHECK1-NEXT:    [[TMP57:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP55]], i32 1, i32 2, i8* [[TMP56]])
183 // CHECK1-NEXT:    store i8* [[TMP57]], i8** [[DOTTASK_RED_]], align 8
184 // CHECK1-NEXT:    [[TMP58:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
185 // CHECK1-NEXT:    [[TMP59:%.*]] = load i32, i32* [[TMP58]], align 4
186 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP59]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
187 // CHECK1-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
188 // CHECK1-NEXT:    [[TMP61:%.*]] = icmp slt i32 [[TMP60]], 0
189 // CHECK1-NEXT:    [[TMP62:%.*]] = select i1 [[TMP61]], i32 [[TMP60]], i32 0
190 // CHECK1-NEXT:    store i32 [[TMP62]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
191 // CHECK1-NEXT:    [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
192 // CHECK1-NEXT:    store i32 [[TMP63]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
193 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
194 // CHECK1:       omp.inner.for.cond:
195 // CHECK1-NEXT:    [[TMP64:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
196 // CHECK1-NEXT:    [[TMP65:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
197 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP64]], [[TMP65]]
198 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
199 // CHECK1:       omp.inner.for.body:
200 // CHECK1-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
201 // CHECK1-NEXT:    switch i32 [[TMP66]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
202 // CHECK1-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
203 // CHECK1-NEXT:    ]
204 // CHECK1:       .omp.sections.case:
205 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
206 // CHECK1-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP67]], align 8
207 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
208 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[TMP68]], align 8
209 // CHECK1-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
210 // CHECK1-NEXT:    [[TMP70:%.*]] = load i8**, i8*** [[TMP]], align 8
211 // CHECK1-NEXT:    store i8** [[TMP70]], i8*** [[TMP69]], align 8
212 // CHECK1-NEXT:    [[TMP71:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
213 // CHECK1-NEXT:    [[TMP72:%.*]] = load i32, i32* [[TMP71]], align 4
214 // CHECK1-NEXT:    [[TMP73:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP72]], 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*)*))
215 // CHECK1-NEXT:    [[TMP74:%.*]] = bitcast i8* [[TMP73]] to %struct.kmp_task_t_with_privates*
216 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP74]], i32 0, i32 0
217 // CHECK1-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP75]], i32 0, i32 0
218 // CHECK1-NEXT:    [[TMP77:%.*]] = load i8*, i8** [[TMP76]], align 8
219 // CHECK1-NEXT:    [[TMP78:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
220 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP77]], i8* align 8 [[TMP78]], i64 24, i1 false)
221 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP74]], i32 0, i32 1
222 // CHECK1-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP79]], i32 0, i32 0
223 // CHECK1-NEXT:    [[TMP81:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
224 // CHECK1-NEXT:    store i8* [[TMP81]], i8** [[TMP80]], align 8
225 // CHECK1-NEXT:    [[TMP82:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
226 // CHECK1-NEXT:    [[TMP83:%.*]] = load i32, i32* [[TMP82]], align 4
227 // CHECK1-NEXT:    [[TMP84:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP83]], i8* [[TMP73]])
228 // CHECK1-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
229 // CHECK1:       .omp.sections.exit:
230 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
231 // CHECK1:       omp.inner.for.inc:
232 // CHECK1-NEXT:    [[TMP85:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
233 // CHECK1-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP85]], 1
234 // CHECK1-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
235 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
236 // CHECK1:       omp.inner.for.end:
237 // CHECK1-NEXT:    [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
238 // CHECK1-NEXT:    [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
239 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP87]])
240 // CHECK1-NEXT:    [[TMP88:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
241 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, i32* [[TMP88]], align 4
242 // CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP89]], i32 1)
243 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
244 // CHECK1-NEXT:    [[TMP91:%.*]] = bitcast i32* [[ARGC]] to i8*
245 // CHECK1-NEXT:    store i8* [[TMP91]], i8** [[TMP90]], align 8
246 // CHECK1-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
247 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP92]], align 8
248 // CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
249 // CHECK1-NEXT:    [[TMP94:%.*]] = inttoptr i64 [[TMP14]] to i8*
250 // CHECK1-NEXT:    store i8* [[TMP94]], i8** [[TMP93]], align 8
251 // CHECK1-NEXT:    [[TMP95:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
252 // CHECK1-NEXT:    [[TMP96:%.*]] = load i32, i32* [[TMP95]], align 4
253 // CHECK1-NEXT:    [[TMP97:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
254 // CHECK1-NEXT:    [[TMP98:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP96]], i32 2, i64 24, i8* [[TMP97]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
255 // CHECK1-NEXT:    switch i32 [[TMP98]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
256 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
257 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
258 // CHECK1-NEXT:    ]
259 // CHECK1:       .omp.reduction.case1:
260 // CHECK1-NEXT:    [[TMP99:%.*]] = load i32, i32* [[TMP2]], align 4
261 // CHECK1-NEXT:    [[TMP100:%.*]] = load i32, i32* [[ARGC]], align 4
262 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP99]], [[TMP100]]
263 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
264 // CHECK1-NEXT:    [[TMP101:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
265 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP101]]
266 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
267 // CHECK1:       omp.arraycpy.body:
268 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
269 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
270 // CHECK1-NEXT:    [[TMP102:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
271 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP102]] to i32
272 // CHECK1-NEXT:    [[TMP103:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
273 // CHECK1-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP103]] to i32
274 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
275 // CHECK1-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
276 // CHECK1-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
277 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
278 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
279 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP101]]
280 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
281 // CHECK1:       omp.arraycpy.done17:
282 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP96]], [8 x i32]* @.gomp_critical_user_.reduction.var)
283 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
284 // CHECK1:       .omp.reduction.case2:
285 // CHECK1-NEXT:    [[TMP104:%.*]] = load i32, i32* [[ARGC]], align 4
286 // CHECK1-NEXT:    [[TMP105:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP104]] monotonic, align 4
287 // CHECK1-NEXT:    [[TMP106:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
288 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP106]]
289 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
290 // CHECK1:       omp.arraycpy.body19:
291 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
292 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
293 // CHECK1-NEXT:    [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
294 // CHECK1-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP107]] to i32
295 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
296 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
297 // CHECK1:       atomic_cont:
298 // CHECK1-NEXT:    [[TMP108:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP113:%.*]], [[ATOMIC_CONT]] ]
299 // CHECK1-NEXT:    store i8 [[TMP108]], i8* [[_TMP23]], align 1
300 // CHECK1-NEXT:    [[TMP109:%.*]] = load i8, i8* [[_TMP23]], align 1
301 // CHECK1-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP109]] to i32
302 // CHECK1-NEXT:    [[TMP110:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
303 // CHECK1-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP110]] to i32
304 // CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
305 // CHECK1-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
306 // CHECK1-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
307 // CHECK1-NEXT:    [[TMP111:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
308 // CHECK1-NEXT:    [[TMP112:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP108]], i8 [[TMP111]] monotonic monotonic, align 1
309 // CHECK1-NEXT:    [[TMP113]] = extractvalue { i8, i1 } [[TMP112]], 0
310 // CHECK1-NEXT:    [[TMP114:%.*]] = extractvalue { i8, i1 } [[TMP112]], 1
311 // CHECK1-NEXT:    br i1 [[TMP114]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
312 // CHECK1:       atomic_exit:
313 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
314 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
315 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP106]]
316 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
317 // CHECK1:       omp.arraycpy.done31:
318 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
319 // CHECK1:       .omp.reduction.default:
320 // CHECK1-NEXT:    [[TMP115:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
321 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP115]])
322 // CHECK1-NEXT:    ret void
323 //
324 //
325 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
326 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
327 // CHECK1-NEXT:  entry:
328 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
329 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
330 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
331 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
332 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
333 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
334 // CHECK1-NEXT:    store i32 0, i32* [[TMP3]], align 8
335 // CHECK1-NEXT:    ret void
336 //
337 //
338 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
339 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
340 // CHECK1-NEXT:  entry:
341 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
342 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
343 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
344 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
345 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
346 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
347 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
348 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
349 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
350 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
351 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
352 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
353 // CHECK1-NEXT:    ret void
354 //
355 //
356 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
357 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
358 // CHECK1-NEXT:  entry:
359 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
360 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
361 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
362 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
363 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
364 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
365 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
366 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
367 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
368 // CHECK1:       omp.arrayinit.body:
369 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
370 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
371 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
372 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
373 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
374 // CHECK1:       omp.arrayinit.done:
375 // CHECK1-NEXT:    ret void
376 //
377 //
378 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
379 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
380 // CHECK1-NEXT:  entry:
381 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
382 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
383 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
384 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
385 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
386 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
387 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
388 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
389 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
390 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
391 // CHECK1:       omp.arraycpy.body:
392 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
393 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
394 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
395 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
396 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
397 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
398 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
399 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
400 // CHECK1-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
401 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
402 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
403 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
404 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
405 // CHECK1:       omp.arraycpy.done4:
406 // CHECK1-NEXT:    ret void
407 //
408 //
409 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
410 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
411 // CHECK1-NEXT:  entry:
412 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
413 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
414 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
415 // CHECK1-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
416 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
417 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
418 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
419 // CHECK1-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
420 // CHECK1-NEXT:    ret void
421 //
422 //
423 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
424 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
425 // CHECK1-NEXT:  entry:
426 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
427 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
428 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
429 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
430 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
431 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
432 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
433 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
434 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
435 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
436 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
437 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
438 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
439 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
440 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
441 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
442 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
443 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
444 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
445 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
446 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
447 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
448 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
449 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
450 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
451 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
452 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
453 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
454 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
455 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
456 // 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
457 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
458 // CHECK1-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
459 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
460 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
461 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
462 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
463 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
464 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
465 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
466 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
467 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
468 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
469 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
470 // CHECK1-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
471 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
472 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
473 // CHECK1-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
474 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
475 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
476 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
477 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
478 // CHECK1-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
479 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
480 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
481 // CHECK1-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
482 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
483 // CHECK1-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
484 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
485 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
486 // CHECK1-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
487 // CHECK1-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
488 // CHECK1-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
489 // CHECK1-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
490 // CHECK1-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
491 // CHECK1-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
492 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
493 // CHECK1-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
494 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
495 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
496 // CHECK1-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
497 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
498 // CHECK1-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
499 // CHECK1-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
500 // CHECK1-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
501 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
502 // CHECK1-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
503 // CHECK1-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
504 // CHECK1-NEXT:    ret i32 0
505 //
506 //
507 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
508 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
509 // CHECK1-NEXT:  entry:
510 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
511 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
512 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
513 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
514 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
515 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
516 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
517 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
518 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
519 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
520 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
521 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
522 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
523 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
524 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
525 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
526 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
527 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
528 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
529 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
530 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
531 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
532 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
533 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
534 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
535 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
536 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
537 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
538 // CHECK1:       omp.arraycpy.body:
539 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
540 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
541 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
542 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
543 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
544 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
545 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
546 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
547 // CHECK1-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
548 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
549 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
550 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
551 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
552 // CHECK1:       omp.arraycpy.done5:
553 // CHECK1-NEXT:    ret void
554 //
555 //
556 // CHECK2-LABEL: define {{[^@]+}}@main
557 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
558 // CHECK2-NEXT:  entry:
559 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
560 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
561 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
562 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
563 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
564 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
565 // CHECK2-NEXT:    store i32* [[ARGC_ADDR]], i32** [[TMP0]], align 8
566 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
567 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
568 // CHECK2-NEXT:    store i8** [[TMP2]], i8*** [[TMP1]], align 8
569 // 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_]])
570 // CHECK2-NEXT:    ret i32 0
571 //
572 //
573 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
574 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
575 // CHECK2-NEXT:  entry:
576 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
577 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
578 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
579 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
580 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
581 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
582 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
583 // CHECK2-NEXT:    [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
584 // CHECK2-NEXT:    [[ARGC:%.*]] = alloca i32, align 4
585 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
586 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
587 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
588 // CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i8*, align 8
589 // CHECK2-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
590 // CHECK2-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
591 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
592 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
593 // CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
594 // CHECK2-NEXT:    [[_TMP23:%.*]] = alloca i8, align 1
595 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
596 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
597 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
598 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
599 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
600 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
601 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
602 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
603 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
604 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
605 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
606 // CHECK2-NEXT:    store i32 0, i32* [[ARGC]], align 4
607 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8**, i8*** [[TMP3]], align 8
608 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP4]], i64 0
609 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
610 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i64 0
611 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP2]], align 4
612 // CHECK2-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
613 // CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
614 // CHECK2-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP3]], align 8
615 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 9
616 // CHECK2-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX2]], align 8
617 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 [[LB_ADD_LEN]]
618 // CHECK2-NEXT:    [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX3]] to i64
619 // CHECK2-NEXT:    [[TMP11:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
620 // CHECK2-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
621 // CHECK2-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
622 // CHECK2-NEXT:    [[TMP14:%.*]] = add nuw i64 [[TMP13]], 1
623 // CHECK2-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
624 // CHECK2-NEXT:    [[TMP16:%.*]] = call i8* @llvm.stacksave()
625 // CHECK2-NEXT:    store i8* [[TMP16]], i8** [[SAVED_STACK]], align 8
626 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP14]], align 16
627 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[__VLA_EXPR0]], align 8
628 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP14]]
629 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP17]]
630 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
631 // CHECK2:       omp.arrayinit.body:
632 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
633 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
634 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
635 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
636 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
637 // CHECK2:       omp.arrayinit.done:
638 // CHECK2-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP3]], align 8
639 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
640 // CHECK2-NEXT:    [[TMP20:%.*]] = ptrtoint i8* [[TMP19]] to i64
641 // CHECK2-NEXT:    [[TMP21:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
642 // CHECK2-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
643 // CHECK2-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
644 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP23]]
645 // CHECK2-NEXT:    store i8** [[_TMP4]], i8*** [[TMP]], align 8
646 // CHECK2-NEXT:    store i8* [[TMP24]], i8** [[_TMP4]], align 8
647 // 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
648 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
649 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[ARGC]] to i8*
650 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
651 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
652 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP2]] to i8*
653 // CHECK2-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
654 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
655 // CHECK2-NEXT:    store i64 4, i64* [[TMP29]], align 8
656 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
657 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP30]], align 8
658 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
659 // CHECK2-NEXT:    store i8* null, i8** [[TMP31]], align 8
660 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
661 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP32]], align 8
662 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
663 // CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i32* [[TMP33]] to i8*
664 // CHECK2-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP34]], i8 0, i64 4, i1 false)
665 // 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
666 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 0
667 // CHECK2-NEXT:    [[TMP36:%.*]] = load i8**, i8*** [[TMP3]], align 8
668 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP36]], i64 0
669 // CHECK2-NEXT:    [[TMP37:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
670 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP37]], i64 0
671 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP2]], align 4
672 // CHECK2-NEXT:    [[TMP39:%.*]] = sext i32 [[TMP38]] to i64
673 // CHECK2-NEXT:    [[LB_ADD_LEN8:%.*]] = add nsw i64 -1, [[TMP39]]
674 // CHECK2-NEXT:    [[TMP40:%.*]] = load i8**, i8*** [[TMP3]], align 8
675 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8*, i8** [[TMP40]], i64 9
676 // CHECK2-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[ARRAYIDX9]], align 8
677 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, i8* [[TMP41]], i64 [[LB_ADD_LEN8]]
678 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP35]], align 8
679 // CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 1
680 // CHECK2-NEXT:    store i8* [[ARRAYIDX7]], i8** [[TMP42]], align 8
681 // CHECK2-NEXT:    [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX10]] to i64
682 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[ARRAYIDX7]] to i64
683 // CHECK2-NEXT:    [[TMP45:%.*]] = sub i64 [[TMP43]], [[TMP44]]
684 // CHECK2-NEXT:    [[TMP46:%.*]] = sdiv exact i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
685 // CHECK2-NEXT:    [[TMP47:%.*]] = add nuw i64 [[TMP46]], 1
686 // CHECK2-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
687 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 2
688 // CHECK2-NEXT:    store i64 [[TMP48]], i64* [[TMP49]], align 8
689 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 3
690 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP50]], align 8
691 // CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 4
692 // CHECK2-NEXT:    store i8* null, i8** [[TMP51]], align 8
693 // CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 5
694 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP52]], align 8
695 // CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 6
696 // CHECK2-NEXT:    store i32 1, i32* [[TMP53]], align 8
697 // CHECK2-NEXT:    [[TMP54:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
698 // CHECK2-NEXT:    [[TMP55:%.*]] = load i32, i32* [[TMP54]], align 4
699 // CHECK2-NEXT:    [[TMP56:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
700 // CHECK2-NEXT:    [[TMP57:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP55]], i32 1, i32 2, i8* [[TMP56]])
701 // CHECK2-NEXT:    store i8* [[TMP57]], i8** [[DOTTASK_RED_]], align 8
702 // CHECK2-NEXT:    [[TMP58:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
703 // CHECK2-NEXT:    [[TMP59:%.*]] = load i32, i32* [[TMP58]], align 4
704 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP59]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
705 // CHECK2-NEXT:    [[TMP60:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
706 // CHECK2-NEXT:    [[TMP61:%.*]] = icmp slt i32 [[TMP60]], 0
707 // CHECK2-NEXT:    [[TMP62:%.*]] = select i1 [[TMP61]], i32 [[TMP60]], i32 0
708 // CHECK2-NEXT:    store i32 [[TMP62]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
709 // CHECK2-NEXT:    [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
710 // CHECK2-NEXT:    store i32 [[TMP63]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
711 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
712 // CHECK2:       omp.inner.for.cond:
713 // CHECK2-NEXT:    [[TMP64:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
714 // CHECK2-NEXT:    [[TMP65:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
715 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP64]], [[TMP65]]
716 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
717 // CHECK2:       omp.inner.for.body:
718 // CHECK2-NEXT:    [[TMP66:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
719 // CHECK2-NEXT:    switch i32 [[TMP66]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
720 // CHECK2-NEXT:    i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
721 // CHECK2-NEXT:    ]
722 // CHECK2:       .omp.sections.case:
723 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
724 // CHECK2-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP67]], align 8
725 // CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
726 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[TMP68]], align 8
727 // CHECK2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
728 // CHECK2-NEXT:    [[TMP70:%.*]] = load i8**, i8*** [[TMP]], align 8
729 // CHECK2-NEXT:    store i8** [[TMP70]], i8*** [[TMP69]], align 8
730 // CHECK2-NEXT:    [[TMP71:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
731 // CHECK2-NEXT:    [[TMP72:%.*]] = load i32, i32* [[TMP71]], align 4
732 // CHECK2-NEXT:    [[TMP73:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP72]], 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*)*))
733 // CHECK2-NEXT:    [[TMP74:%.*]] = bitcast i8* [[TMP73]] to %struct.kmp_task_t_with_privates*
734 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP74]], i32 0, i32 0
735 // CHECK2-NEXT:    [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP75]], i32 0, i32 0
736 // CHECK2-NEXT:    [[TMP77:%.*]] = load i8*, i8** [[TMP76]], align 8
737 // CHECK2-NEXT:    [[TMP78:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
738 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP77]], i8* align 8 [[TMP78]], i64 24, i1 false)
739 // CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP74]], i32 0, i32 1
740 // CHECK2-NEXT:    [[TMP80:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP79]], i32 0, i32 0
741 // CHECK2-NEXT:    [[TMP81:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
742 // CHECK2-NEXT:    store i8* [[TMP81]], i8** [[TMP80]], align 8
743 // CHECK2-NEXT:    [[TMP82:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
744 // CHECK2-NEXT:    [[TMP83:%.*]] = load i32, i32* [[TMP82]], align 4
745 // CHECK2-NEXT:    [[TMP84:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP83]], i8* [[TMP73]])
746 // CHECK2-NEXT:    br label [[DOTOMP_SECTIONS_EXIT]]
747 // CHECK2:       .omp.sections.exit:
748 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
749 // CHECK2:       omp.inner.for.inc:
750 // CHECK2-NEXT:    [[TMP85:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
751 // CHECK2-NEXT:    [[INC:%.*]] = add nsw i32 [[TMP85]], 1
752 // CHECK2-NEXT:    store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
753 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
754 // CHECK2:       omp.inner.for.end:
755 // CHECK2-NEXT:    [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
756 // CHECK2-NEXT:    [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
757 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP87]])
758 // CHECK2-NEXT:    [[TMP88:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
759 // CHECK2-NEXT:    [[TMP89:%.*]] = load i32, i32* [[TMP88]], align 4
760 // CHECK2-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP89]], i32 1)
761 // CHECK2-NEXT:    [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
762 // CHECK2-NEXT:    [[TMP91:%.*]] = bitcast i32* [[ARGC]] to i8*
763 // CHECK2-NEXT:    store i8* [[TMP91]], i8** [[TMP90]], align 8
764 // CHECK2-NEXT:    [[TMP92:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
765 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP92]], align 8
766 // CHECK2-NEXT:    [[TMP93:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
767 // CHECK2-NEXT:    [[TMP94:%.*]] = inttoptr i64 [[TMP14]] to i8*
768 // CHECK2-NEXT:    store i8* [[TMP94]], i8** [[TMP93]], align 8
769 // CHECK2-NEXT:    [[TMP95:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
770 // CHECK2-NEXT:    [[TMP96:%.*]] = load i32, i32* [[TMP95]], align 4
771 // CHECK2-NEXT:    [[TMP97:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
772 // CHECK2-NEXT:    [[TMP98:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP96]], i32 2, i64 24, i8* [[TMP97]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
773 // CHECK2-NEXT:    switch i32 [[TMP98]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
774 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
775 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
776 // CHECK2-NEXT:    ]
777 // CHECK2:       .omp.reduction.case1:
778 // CHECK2-NEXT:    [[TMP99:%.*]] = load i32, i32* [[TMP2]], align 4
779 // CHECK2-NEXT:    [[TMP100:%.*]] = load i32, i32* [[ARGC]], align 4
780 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP99]], [[TMP100]]
781 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
782 // CHECK2-NEXT:    [[TMP101:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
783 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP101]]
784 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
785 // CHECK2:       omp.arraycpy.body:
786 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
787 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
788 // CHECK2-NEXT:    [[TMP102:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
789 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP102]] to i32
790 // CHECK2-NEXT:    [[TMP103:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
791 // CHECK2-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP103]] to i32
792 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
793 // CHECK2-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
794 // CHECK2-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
795 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
796 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
797 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP101]]
798 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
799 // CHECK2:       omp.arraycpy.done17:
800 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP96]], [8 x i32]* @.gomp_critical_user_.reduction.var)
801 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
802 // CHECK2:       .omp.reduction.case2:
803 // CHECK2-NEXT:    [[TMP104:%.*]] = load i32, i32* [[ARGC]], align 4
804 // CHECK2-NEXT:    [[TMP105:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP104]] monotonic, align 4
805 // CHECK2-NEXT:    [[TMP106:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
806 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP106]]
807 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
808 // CHECK2:       omp.arraycpy.body19:
809 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
810 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
811 // CHECK2-NEXT:    [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
812 // CHECK2-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP107]] to i32
813 // CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
814 // CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
815 // CHECK2:       atomic_cont:
816 // CHECK2-NEXT:    [[TMP108:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP113:%.*]], [[ATOMIC_CONT]] ]
817 // CHECK2-NEXT:    store i8 [[TMP108]], i8* [[_TMP23]], align 1
818 // CHECK2-NEXT:    [[TMP109:%.*]] = load i8, i8* [[_TMP23]], align 1
819 // CHECK2-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP109]] to i32
820 // CHECK2-NEXT:    [[TMP110:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
821 // CHECK2-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP110]] to i32
822 // CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
823 // CHECK2-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
824 // CHECK2-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
825 // CHECK2-NEXT:    [[TMP111:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
826 // CHECK2-NEXT:    [[TMP112:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP108]], i8 [[TMP111]] monotonic monotonic, align 1
827 // CHECK2-NEXT:    [[TMP113]] = extractvalue { i8, i1 } [[TMP112]], 0
828 // CHECK2-NEXT:    [[TMP114:%.*]] = extractvalue { i8, i1 } [[TMP112]], 1
829 // CHECK2-NEXT:    br i1 [[TMP114]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
830 // CHECK2:       atomic_exit:
831 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
832 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
833 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP106]]
834 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
835 // CHECK2:       omp.arraycpy.done31:
836 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
837 // CHECK2:       .omp.reduction.default:
838 // CHECK2-NEXT:    [[TMP115:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
839 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP115]])
840 // CHECK2-NEXT:    ret void
841 //
842 //
843 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
844 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
845 // CHECK2-NEXT:  entry:
846 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
847 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
848 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
849 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
850 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
851 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
852 // CHECK2-NEXT:    store i32 0, i32* [[TMP3]], align 8
853 // CHECK2-NEXT:    ret void
854 //
855 //
856 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
857 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
858 // CHECK2-NEXT:  entry:
859 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
860 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
861 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
862 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
863 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
864 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
865 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
866 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
867 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
868 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
869 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
870 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
871 // CHECK2-NEXT:    ret void
872 //
873 //
874 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
875 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
876 // CHECK2-NEXT:  entry:
877 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
878 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
879 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
880 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
881 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
882 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
883 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
884 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
885 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
886 // CHECK2:       omp.arrayinit.body:
887 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
888 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
889 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
890 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
891 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
892 // CHECK2:       omp.arrayinit.done:
893 // CHECK2-NEXT:    ret void
894 //
895 //
896 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
897 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
898 // CHECK2-NEXT:  entry:
899 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
900 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
901 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
902 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
903 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
904 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
905 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
906 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
907 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
908 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
909 // CHECK2:       omp.arraycpy.body:
910 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
911 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
912 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
913 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
914 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
915 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
916 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
917 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
918 // CHECK2-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
919 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
920 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
921 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
922 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
923 // CHECK2:       omp.arraycpy.done4:
924 // CHECK2-NEXT:    ret void
925 //
926 //
927 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
928 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
929 // CHECK2-NEXT:  entry:
930 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
931 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
932 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
933 // CHECK2-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
934 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
935 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
936 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
937 // CHECK2-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
938 // CHECK2-NEXT:    ret void
939 //
940 //
941 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
942 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
943 // CHECK2-NEXT:  entry:
944 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
945 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
946 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
947 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
948 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
949 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
950 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
951 // CHECK2-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
952 // CHECK2-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
953 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
954 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
955 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
956 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
957 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
958 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
959 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
960 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
961 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
962 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
963 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
964 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
965 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
966 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
967 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
968 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
969 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
970 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
971 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
972 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
973 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
974 // 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
975 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
976 // CHECK2-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
977 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
978 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
979 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
980 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
981 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
982 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
983 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
984 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
985 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
986 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
987 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
988 // CHECK2-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
989 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
990 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
991 // CHECK2-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
992 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
993 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
994 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
995 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
996 // CHECK2-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
997 // CHECK2-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
998 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
999 // CHECK2-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
1000 // CHECK2-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
1001 // CHECK2-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
1002 // CHECK2-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
1003 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
1004 // CHECK2-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
1005 // CHECK2-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
1006 // CHECK2-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1007 // CHECK2-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
1008 // CHECK2-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1009 // CHECK2-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
1010 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
1011 // CHECK2-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1012 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
1013 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1014 // CHECK2-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1015 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1016 // CHECK2-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1017 // CHECK2-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1018 // CHECK2-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1019 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1020 // CHECK2-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1021 // CHECK2-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1022 // CHECK2-NEXT:    ret i32 0
1023 //
1024 //
1025 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1026 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1027 // CHECK2-NEXT:  entry:
1028 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
1029 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
1030 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1031 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1032 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1033 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1034 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1035 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1036 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1037 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1038 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1039 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1040 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1041 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1042 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1043 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1044 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1045 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1046 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1047 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1048 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1049 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1050 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1051 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1052 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
1053 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1054 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1055 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1056 // CHECK2:       omp.arraycpy.body:
1057 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1058 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1059 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1060 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1061 // CHECK2-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1062 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1063 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1064 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1065 // CHECK2-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1066 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1067 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1068 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1069 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1070 // CHECK2:       omp.arraycpy.done5:
1071 // CHECK2-NEXT:    ret void
1072 //
1073