1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-linux -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -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 -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -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 master 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:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
176 // CHECK1-NEXT:    [[TMP59:%.*]] = load i32, i32* [[TMP58]], align 4
177 // CHECK1-NEXT:    [[TMP60:%.*]] = call i32 @__kmpc_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP59]])
178 // CHECK1-NEXT:    [[TMP61:%.*]] = icmp ne i32 [[TMP60]], 0
179 // CHECK1-NEXT:    br i1 [[TMP61]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_END:%.*]]
180 // CHECK1:       omp_if.then:
181 // CHECK1-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
182 // CHECK1-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP62]], align 8
183 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
184 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[TMP63]], align 8
185 // CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
186 // CHECK1-NEXT:    [[TMP65:%.*]] = load i8**, i8*** [[TMP]], align 8
187 // CHECK1-NEXT:    store i8** [[TMP65]], i8*** [[TMP64]], align 8
188 // CHECK1-NEXT:    [[TMP66:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
189 // CHECK1-NEXT:    [[TMP67:%.*]] = load i32, i32* [[TMP66]], align 4
190 // CHECK1-NEXT:    [[TMP68:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP67]], 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*)*))
191 // CHECK1-NEXT:    [[TMP69:%.*]] = bitcast i8* [[TMP68]] to %struct.kmp_task_t_with_privates*
192 // CHECK1-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP69]], i32 0, i32 0
193 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP70]], i32 0, i32 0
194 // CHECK1-NEXT:    [[TMP72:%.*]] = load i8*, i8** [[TMP71]], align 8
195 // CHECK1-NEXT:    [[TMP73:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
196 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP72]], i8* align 8 [[TMP73]], i64 24, i1 false)
197 // CHECK1-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP69]], i32 0, i32 1
198 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP74]], i32 0, i32 0
199 // CHECK1-NEXT:    [[TMP76:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
200 // CHECK1-NEXT:    store i8* [[TMP76]], i8** [[TMP75]], align 8
201 // CHECK1-NEXT:    [[TMP77:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
202 // CHECK1-NEXT:    [[TMP78:%.*]] = load i32, i32* [[TMP77]], align 4
203 // CHECK1-NEXT:    [[TMP79:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP78]], i8* [[TMP68]])
204 // CHECK1-NEXT:    call void @__kmpc_end_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP59]])
205 // CHECK1-NEXT:    br label [[OMP_IF_END]]
206 // CHECK1:       omp_if.end:
207 // CHECK1-NEXT:    [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
208 // CHECK1-NEXT:    [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
209 // CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i32 0)
210 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
211 // CHECK1-NEXT:    [[TMP83:%.*]] = bitcast i32* [[ARGC]] to i8*
212 // CHECK1-NEXT:    store i8* [[TMP83]], i8** [[TMP82]], align 8
213 // CHECK1-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
214 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP84]], align 8
215 // CHECK1-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
216 // CHECK1-NEXT:    [[TMP86:%.*]] = inttoptr i64 [[TMP14]] to i8*
217 // CHECK1-NEXT:    store i8* [[TMP86]], i8** [[TMP85]], align 8
218 // CHECK1-NEXT:    [[TMP87:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
219 // CHECK1-NEXT:    [[TMP88:%.*]] = load i32, i32* [[TMP87]], align 4
220 // CHECK1-NEXT:    [[TMP89:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
221 // CHECK1-NEXT:    [[TMP90:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP88]], i32 2, i64 24, i8* [[TMP89]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
222 // CHECK1-NEXT:    switch i32 [[TMP90]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
223 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
224 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
225 // CHECK1-NEXT:    ]
226 // CHECK1:       .omp.reduction.case1:
227 // CHECK1-NEXT:    [[TMP91:%.*]] = load i32, i32* [[TMP2]], align 4
228 // CHECK1-NEXT:    [[TMP92:%.*]] = load i32, i32* [[ARGC]], align 4
229 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP91]], [[TMP92]]
230 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
231 // CHECK1-NEXT:    [[TMP93:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
232 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP93]]
233 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
234 // CHECK1:       omp.arraycpy.body:
235 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
236 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
237 // CHECK1-NEXT:    [[TMP94:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
238 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP94]] to i32
239 // CHECK1-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
240 // CHECK1-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP95]] to i32
241 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
242 // CHECK1-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
243 // CHECK1-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
244 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
245 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
246 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP93]]
247 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
248 // CHECK1:       omp.arraycpy.done17:
249 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP88]], [8 x i32]* @.gomp_critical_user_.reduction.var)
250 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
251 // CHECK1:       .omp.reduction.case2:
252 // CHECK1-NEXT:    [[TMP96:%.*]] = load i32, i32* [[ARGC]], align 4
253 // CHECK1-NEXT:    [[TMP97:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP96]] monotonic, align 4
254 // CHECK1-NEXT:    [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
255 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP98]]
256 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
257 // CHECK1:       omp.arraycpy.body19:
258 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
259 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
260 // CHECK1-NEXT:    [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
261 // CHECK1-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP99]] to i32
262 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
263 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
264 // CHECK1:       atomic_cont:
265 // CHECK1-NEXT:    [[TMP100:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP105:%.*]], [[ATOMIC_CONT]] ]
266 // CHECK1-NEXT:    store i8 [[TMP100]], i8* [[_TMP23]], align 1
267 // CHECK1-NEXT:    [[TMP101:%.*]] = load i8, i8* [[_TMP23]], align 1
268 // CHECK1-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP101]] to i32
269 // CHECK1-NEXT:    [[TMP102:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
270 // CHECK1-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP102]] to i32
271 // CHECK1-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
272 // CHECK1-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
273 // CHECK1-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
274 // CHECK1-NEXT:    [[TMP103:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
275 // CHECK1-NEXT:    [[TMP104:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP100]], i8 [[TMP103]] monotonic monotonic, align 1
276 // CHECK1-NEXT:    [[TMP105]] = extractvalue { i8, i1 } [[TMP104]], 0
277 // CHECK1-NEXT:    [[TMP106:%.*]] = extractvalue { i8, i1 } [[TMP104]], 1
278 // CHECK1-NEXT:    br i1 [[TMP106]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
279 // CHECK1:       atomic_exit:
280 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
281 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
282 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP98]]
283 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
284 // CHECK1:       omp.arraycpy.done31:
285 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
286 // CHECK1:       .omp.reduction.default:
287 // CHECK1-NEXT:    [[TMP107:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
288 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP107]])
289 // CHECK1-NEXT:    ret void
290 //
291 //
292 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
293 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
294 // CHECK1-NEXT:  entry:
295 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
296 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
297 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
298 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
299 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
300 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
301 // CHECK1-NEXT:    store i32 0, i32* [[TMP3]], align 8
302 // CHECK1-NEXT:    ret void
303 //
304 //
305 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
306 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
307 // CHECK1-NEXT:  entry:
308 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
309 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
310 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
311 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
312 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
313 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
314 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
315 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
316 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
317 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
318 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
319 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
320 // CHECK1-NEXT:    ret void
321 //
322 //
323 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
324 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
325 // CHECK1-NEXT:  entry:
326 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
327 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
328 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
329 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
330 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
331 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
332 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
333 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
334 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
335 // CHECK1:       omp.arrayinit.body:
336 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
337 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
338 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
339 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
340 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
341 // CHECK1:       omp.arrayinit.done:
342 // CHECK1-NEXT:    ret void
343 //
344 //
345 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
346 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
347 // CHECK1-NEXT:  entry:
348 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
349 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
350 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
351 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
352 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
353 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
354 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
355 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
356 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
357 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
358 // CHECK1:       omp.arraycpy.body:
359 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
360 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
361 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
362 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
363 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
364 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
365 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
366 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
367 // CHECK1-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
368 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
369 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
370 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
371 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
372 // CHECK1:       omp.arraycpy.done4:
373 // CHECK1-NEXT:    ret void
374 //
375 //
376 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
377 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
378 // CHECK1-NEXT:  entry:
379 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
380 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
381 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
382 // CHECK1-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
383 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
384 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
385 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
386 // CHECK1-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
387 // CHECK1-NEXT:    ret void
388 //
389 //
390 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
391 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
392 // CHECK1-NEXT:  entry:
393 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
394 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
395 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
396 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
397 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
398 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
399 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
400 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
401 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
402 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
403 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
404 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
405 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
406 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
407 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
408 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
409 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
410 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
411 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
412 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
413 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
414 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
415 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
416 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
417 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
418 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
419 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
420 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
421 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
422 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
423 // 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
424 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
425 // CHECK1-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
426 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
427 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
428 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
429 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
430 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
431 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
432 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
433 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
434 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
435 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
436 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
437 // CHECK1-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
438 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
439 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
440 // CHECK1-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
441 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
442 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
443 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
444 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
445 // CHECK1-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
446 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
447 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
448 // CHECK1-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
449 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
450 // CHECK1-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
451 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
452 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
453 // CHECK1-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
454 // CHECK1-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
455 // CHECK1-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
456 // CHECK1-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
457 // CHECK1-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
458 // CHECK1-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
459 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
460 // CHECK1-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
461 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
462 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
463 // CHECK1-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
464 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
465 // CHECK1-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
466 // CHECK1-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
467 // CHECK1-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
468 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
469 // CHECK1-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
470 // CHECK1-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
471 // CHECK1-NEXT:    ret i32 0
472 //
473 //
474 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
475 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
476 // CHECK1-NEXT:  entry:
477 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
478 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
479 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
480 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
481 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
482 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
483 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
484 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
485 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
486 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
487 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
488 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
489 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
490 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
491 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
492 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
493 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
494 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
495 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
496 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
497 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
498 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
499 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
500 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
501 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
502 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
503 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
504 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
505 // CHECK1:       omp.arraycpy.body:
506 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
507 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
508 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
509 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
510 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
511 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
512 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
513 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
514 // CHECK1-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
515 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
516 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
517 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
518 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
519 // CHECK1:       omp.arraycpy.done5:
520 // CHECK1-NEXT:    ret void
521 //
522 //
523 // CHECK2-LABEL: define {{[^@]+}}@main
524 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
525 // CHECK2-NEXT:  entry:
526 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
527 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
528 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
529 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
530 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
531 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
532 // CHECK2-NEXT:    store i32* [[ARGC_ADDR]], i32** [[TMP0]], align 8
533 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
534 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
535 // CHECK2-NEXT:    store i8** [[TMP2]], i8*** [[TMP1]], align 8
536 // 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_]])
537 // CHECK2-NEXT:    ret i32 0
538 //
539 //
540 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
541 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
542 // CHECK2-NEXT:  entry:
543 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
544 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
545 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
546 // CHECK2-NEXT:    [[ARGC:%.*]] = alloca i32, align 4
547 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
548 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
549 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
550 // CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i8*, align 8
551 // CHECK2-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
552 // CHECK2-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
553 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
554 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
555 // CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
556 // CHECK2-NEXT:    [[_TMP23:%.*]] = alloca i8, align 1
557 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
558 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
559 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
560 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
561 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
562 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
563 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP0]], i32 0, i32 1
564 // CHECK2-NEXT:    store i32 0, i32* [[ARGC]], align 4
565 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8**, i8*** [[TMP3]], align 8
566 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP4]], i64 0
567 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
568 // CHECK2-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[TMP5]], i64 0
569 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP2]], align 4
570 // CHECK2-NEXT:    [[TMP7:%.*]] = sext i32 [[TMP6]] to i64
571 // CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP7]]
572 // CHECK2-NEXT:    [[TMP8:%.*]] = load i8**, i8*** [[TMP3]], align 8
573 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8*, i8** [[TMP8]], i64 9
574 // CHECK2-NEXT:    [[TMP9:%.*]] = load i8*, i8** [[ARRAYIDX2]], align 8
575 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, i8* [[TMP9]], i64 [[LB_ADD_LEN]]
576 // CHECK2-NEXT:    [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX3]] to i64
577 // CHECK2-NEXT:    [[TMP11:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
578 // CHECK2-NEXT:    [[TMP12:%.*]] = sub i64 [[TMP10]], [[TMP11]]
579 // CHECK2-NEXT:    [[TMP13:%.*]] = sdiv exact i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
580 // CHECK2-NEXT:    [[TMP14:%.*]] = add nuw i64 [[TMP13]], 1
581 // CHECK2-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
582 // CHECK2-NEXT:    [[TMP16:%.*]] = call i8* @llvm.stacksave()
583 // CHECK2-NEXT:    store i8* [[TMP16]], i8** [[SAVED_STACK]], align 8
584 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP14]], align 16
585 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[__VLA_EXPR0]], align 8
586 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP14]]
587 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP17]]
588 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
589 // CHECK2:       omp.arrayinit.body:
590 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
591 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
592 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
593 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP17]]
594 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
595 // CHECK2:       omp.arrayinit.done:
596 // CHECK2-NEXT:    [[TMP18:%.*]] = load i8**, i8*** [[TMP3]], align 8
597 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP18]], align 8
598 // CHECK2-NEXT:    [[TMP20:%.*]] = ptrtoint i8* [[TMP19]] to i64
599 // CHECK2-NEXT:    [[TMP21:%.*]] = ptrtoint i8* [[ARRAYIDX1]] to i64
600 // CHECK2-NEXT:    [[TMP22:%.*]] = sub i64 [[TMP20]], [[TMP21]]
601 // CHECK2-NEXT:    [[TMP23:%.*]] = sdiv exact i64 [[TMP22]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
602 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP23]]
603 // CHECK2-NEXT:    store i8** [[_TMP4]], i8*** [[TMP]], align 8
604 // CHECK2-NEXT:    store i8* [[TMP24]], i8** [[_TMP4]], align 8
605 // 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
606 // CHECK2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
607 // CHECK2-NEXT:    [[TMP26:%.*]] = bitcast i32* [[ARGC]] to i8*
608 // CHECK2-NEXT:    store i8* [[TMP26]], i8** [[TMP25]], align 8
609 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
610 // CHECK2-NEXT:    [[TMP28:%.*]] = bitcast i32* [[TMP2]] to i8*
611 // CHECK2-NEXT:    store i8* [[TMP28]], i8** [[TMP27]], align 8
612 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
613 // CHECK2-NEXT:    store i64 4, i64* [[TMP29]], align 8
614 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
615 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP30]], align 8
616 // CHECK2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
617 // CHECK2-NEXT:    store i8* null, i8** [[TMP31]], align 8
618 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
619 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP32]], align 8
620 // CHECK2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
621 // CHECK2-NEXT:    [[TMP34:%.*]] = bitcast i32* [[TMP33]] to i8*
622 // CHECK2-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP34]], i8 0, i64 4, i1 false)
623 // 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
624 // CHECK2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 0
625 // CHECK2-NEXT:    [[TMP36:%.*]] = load i8**, i8*** [[TMP3]], align 8
626 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds i8*, i8** [[TMP36]], i64 0
627 // CHECK2-NEXT:    [[TMP37:%.*]] = load i8*, i8** [[ARRAYIDX6]], align 8
628 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8, i8* [[TMP37]], i64 0
629 // CHECK2-NEXT:    [[TMP38:%.*]] = load i32, i32* [[TMP2]], align 4
630 // CHECK2-NEXT:    [[TMP39:%.*]] = sext i32 [[TMP38]] to i64
631 // CHECK2-NEXT:    [[LB_ADD_LEN8:%.*]] = add nsw i64 -1, [[TMP39]]
632 // CHECK2-NEXT:    [[TMP40:%.*]] = load i8**, i8*** [[TMP3]], align 8
633 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds i8*, i8** [[TMP40]], i64 9
634 // CHECK2-NEXT:    [[TMP41:%.*]] = load i8*, i8** [[ARRAYIDX9]], align 8
635 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8, i8* [[TMP41]], i64 [[LB_ADD_LEN8]]
636 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP35]], align 8
637 // CHECK2-NEXT:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 1
638 // CHECK2-NEXT:    store i8* [[ARRAYIDX7]], i8** [[TMP42]], align 8
639 // CHECK2-NEXT:    [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX10]] to i64
640 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[ARRAYIDX7]] to i64
641 // CHECK2-NEXT:    [[TMP45:%.*]] = sub i64 [[TMP43]], [[TMP44]]
642 // CHECK2-NEXT:    [[TMP46:%.*]] = sdiv exact i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
643 // CHECK2-NEXT:    [[TMP47:%.*]] = add nuw i64 [[TMP46]], 1
644 // CHECK2-NEXT:    [[TMP48:%.*]] = mul nuw i64 [[TMP47]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
645 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 2
646 // CHECK2-NEXT:    store i64 [[TMP48]], i64* [[TMP49]], align 8
647 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 3
648 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP50]], align 8
649 // CHECK2-NEXT:    [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 4
650 // CHECK2-NEXT:    store i8* null, i8** [[TMP51]], align 8
651 // CHECK2-NEXT:    [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 5
652 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP52]], align 8
653 // CHECK2-NEXT:    [[TMP53:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_5]], i32 0, i32 6
654 // CHECK2-NEXT:    store i32 1, i32* [[TMP53]], align 8
655 // CHECK2-NEXT:    [[TMP54:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
656 // CHECK2-NEXT:    [[TMP55:%.*]] = load i32, i32* [[TMP54]], align 4
657 // CHECK2-NEXT:    [[TMP56:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
658 // CHECK2-NEXT:    [[TMP57:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP55]], i32 0, i32 2, i8* [[TMP56]])
659 // CHECK2-NEXT:    store i8* [[TMP57]], i8** [[DOTTASK_RED_]], align 8
660 // CHECK2-NEXT:    [[TMP58:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
661 // CHECK2-NEXT:    [[TMP59:%.*]] = load i32, i32* [[TMP58]], align 4
662 // CHECK2-NEXT:    [[TMP60:%.*]] = call i32 @__kmpc_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP59]])
663 // CHECK2-NEXT:    [[TMP61:%.*]] = icmp ne i32 [[TMP60]], 0
664 // CHECK2-NEXT:    br i1 [[TMP61]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_END:%.*]]
665 // CHECK2:       omp_if.then:
666 // CHECK2-NEXT:    [[TMP62:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 0
667 // CHECK2-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP62]], align 8
668 // CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 1
669 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[TMP63]], align 8
670 // CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[AGG_CAPTURED]], i32 0, i32 2
671 // CHECK2-NEXT:    [[TMP65:%.*]] = load i8**, i8*** [[TMP]], align 8
672 // CHECK2-NEXT:    store i8** [[TMP65]], i8*** [[TMP64]], align 8
673 // CHECK2-NEXT:    [[TMP66:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
674 // CHECK2-NEXT:    [[TMP67:%.*]] = load i32, i32* [[TMP66]], align 4
675 // CHECK2-NEXT:    [[TMP68:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP67]], 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*)*))
676 // CHECK2-NEXT:    [[TMP69:%.*]] = bitcast i8* [[TMP68]] to %struct.kmp_task_t_with_privates*
677 // CHECK2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP69]], i32 0, i32 0
678 // CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP70]], i32 0, i32 0
679 // CHECK2-NEXT:    [[TMP72:%.*]] = load i8*, i8** [[TMP71]], align 8
680 // CHECK2-NEXT:    [[TMP73:%.*]] = bitcast %struct.anon.0* [[AGG_CAPTURED]] to i8*
681 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP72]], i8* align 8 [[TMP73]], i64 24, i1 false)
682 // CHECK2-NEXT:    [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP69]], i32 0, i32 1
683 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP74]], i32 0, i32 0
684 // CHECK2-NEXT:    [[TMP76:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
685 // CHECK2-NEXT:    store i8* [[TMP76]], i8** [[TMP75]], align 8
686 // CHECK2-NEXT:    [[TMP77:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
687 // CHECK2-NEXT:    [[TMP78:%.*]] = load i32, i32* [[TMP77]], align 4
688 // CHECK2-NEXT:    [[TMP79:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP78]], i8* [[TMP68]])
689 // CHECK2-NEXT:    call void @__kmpc_end_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP59]])
690 // CHECK2-NEXT:    br label [[OMP_IF_END]]
691 // CHECK2:       omp_if.end:
692 // CHECK2-NEXT:    [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
693 // CHECK2-NEXT:    [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
694 // CHECK2-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i32 0)
695 // CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
696 // CHECK2-NEXT:    [[TMP83:%.*]] = bitcast i32* [[ARGC]] to i8*
697 // CHECK2-NEXT:    store i8* [[TMP83]], i8** [[TMP82]], align 8
698 // CHECK2-NEXT:    [[TMP84:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
699 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP84]], align 8
700 // CHECK2-NEXT:    [[TMP85:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
701 // CHECK2-NEXT:    [[TMP86:%.*]] = inttoptr i64 [[TMP14]] to i8*
702 // CHECK2-NEXT:    store i8* [[TMP86]], i8** [[TMP85]], align 8
703 // CHECK2-NEXT:    [[TMP87:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
704 // CHECK2-NEXT:    [[TMP88:%.*]] = load i32, i32* [[TMP87]], align 4
705 // CHECK2-NEXT:    [[TMP89:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
706 // CHECK2-NEXT:    [[TMP90:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP88]], i32 2, i64 24, i8* [[TMP89]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
707 // CHECK2-NEXT:    switch i32 [[TMP90]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
708 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
709 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
710 // CHECK2-NEXT:    ]
711 // CHECK2:       .omp.reduction.case1:
712 // CHECK2-NEXT:    [[TMP91:%.*]] = load i32, i32* [[TMP2]], align 4
713 // CHECK2-NEXT:    [[TMP92:%.*]] = load i32, i32* [[ARGC]], align 4
714 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP91]], [[TMP92]]
715 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP2]], align 4
716 // CHECK2-NEXT:    [[TMP93:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
717 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP93]]
718 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE17:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
719 // CHECK2:       omp.arraycpy.body:
720 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
721 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST11:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT15:%.*]], [[OMP_ARRAYCPY_BODY]] ]
722 // CHECK2-NEXT:    [[TMP94:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
723 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP94]] to i32
724 // CHECK2-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
725 // CHECK2-NEXT:    [[CONV12:%.*]] = sext i8 [[TMP95]] to i32
726 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i32 [[CONV]], [[CONV12]]
727 // CHECK2-NEXT:    [[CONV14:%.*]] = trunc i32 [[ADD13]] to i8
728 // CHECK2-NEXT:    store i8 [[CONV14]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], align 1
729 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT15]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST11]], i32 1
730 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
731 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE16:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT15]], [[TMP93]]
732 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE16]], label [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_BODY]]
733 // CHECK2:       omp.arraycpy.done17:
734 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP88]], [8 x i32]* @.gomp_critical_user_.reduction.var)
735 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
736 // CHECK2:       .omp.reduction.case2:
737 // CHECK2-NEXT:    [[TMP96:%.*]] = load i32, i32* [[ARGC]], align 4
738 // CHECK2-NEXT:    [[TMP97:%.*]] = atomicrmw add i32* [[TMP2]], i32 [[TMP96]] monotonic, align 4
739 // CHECK2-NEXT:    [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX1]], i64 [[TMP14]]
740 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY18:%.*]] = icmp eq i8* [[ARRAYIDX1]], [[TMP98]]
741 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY18]], label [[OMP_ARRAYCPY_DONE31:%.*]], label [[OMP_ARRAYCPY_BODY19:%.*]]
742 // CHECK2:       omp.arraycpy.body19:
743 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST20:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT29:%.*]], [[ATOMIC_EXIT:%.*]] ]
744 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST21:%.*]] = phi i8* [ [[ARRAYIDX1]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT28:%.*]], [[ATOMIC_EXIT]] ]
745 // CHECK2-NEXT:    [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
746 // CHECK2-NEXT:    [[CONV22:%.*]] = sext i8 [[TMP99]] to i32
747 // CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]] monotonic, align 1
748 // CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
749 // CHECK2:       atomic_cont:
750 // CHECK2-NEXT:    [[TMP100:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY19]] ], [ [[TMP105:%.*]], [[ATOMIC_CONT]] ]
751 // CHECK2-NEXT:    store i8 [[TMP100]], i8* [[_TMP23]], align 1
752 // CHECK2-NEXT:    [[TMP101:%.*]] = load i8, i8* [[_TMP23]], align 1
753 // CHECK2-NEXT:    [[CONV24:%.*]] = sext i8 [[TMP101]] to i32
754 // CHECK2-NEXT:    [[TMP102:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], align 1
755 // CHECK2-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP102]] to i32
756 // CHECK2-NEXT:    [[ADD26:%.*]] = add nsw i32 [[CONV24]], [[CONV25]]
757 // CHECK2-NEXT:    [[CONV27:%.*]] = trunc i32 [[ADD26]] to i8
758 // CHECK2-NEXT:    store i8 [[CONV27]], i8* [[ATOMIC_TEMP]], align 1
759 // CHECK2-NEXT:    [[TMP103:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
760 // CHECK2-NEXT:    [[TMP104:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i8 [[TMP100]], i8 [[TMP103]] monotonic monotonic, align 1
761 // CHECK2-NEXT:    [[TMP105]] = extractvalue { i8, i1 } [[TMP104]], 0
762 // CHECK2-NEXT:    [[TMP106:%.*]] = extractvalue { i8, i1 } [[TMP104]], 1
763 // CHECK2-NEXT:    br i1 [[TMP106]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
764 // CHECK2:       atomic_exit:
765 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT28]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST21]], i32 1
766 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST20]], i32 1
767 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE30:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT28]], [[TMP98]]
768 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE30]], label [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_BODY19]]
769 // CHECK2:       omp.arraycpy.done31:
770 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
771 // CHECK2:       .omp.reduction.default:
772 // CHECK2-NEXT:    [[TMP107:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
773 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP107]])
774 // CHECK2-NEXT:    ret void
775 //
776 //
777 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
778 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
779 // CHECK2-NEXT:  entry:
780 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
781 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
782 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
783 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
784 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
785 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
786 // CHECK2-NEXT:    store i32 0, i32* [[TMP3]], align 8
787 // CHECK2-NEXT:    ret void
788 //
789 //
790 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
791 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[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:%.*]] = bitcast i8* [[TMP2]] to i32*
799 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
800 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
801 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
802 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
803 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
804 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
805 // CHECK2-NEXT:    ret void
806 //
807 //
808 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
809 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
810 // CHECK2-NEXT:  entry:
811 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
812 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
813 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
814 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
815 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
816 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
817 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
818 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
819 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
820 // CHECK2:       omp.arrayinit.body:
821 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
822 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
823 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
824 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
825 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
826 // CHECK2:       omp.arrayinit.done:
827 // CHECK2-NEXT:    ret void
828 //
829 //
830 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
831 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
832 // CHECK2-NEXT:  entry:
833 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
834 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
835 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
836 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
837 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
838 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
839 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
840 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
841 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
842 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
843 // CHECK2:       omp.arraycpy.body:
844 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
845 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
846 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
847 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
848 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
849 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
850 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
851 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
852 // CHECK2-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
853 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
854 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
855 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
856 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
857 // CHECK2:       omp.arraycpy.done4:
858 // CHECK2-NEXT:    ret void
859 //
860 //
861 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
862 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
863 // CHECK2-NEXT:  entry:
864 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
865 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
866 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
867 // CHECK2-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
868 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
869 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
870 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
871 // CHECK2-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
872 // CHECK2-NEXT:    ret void
873 //
874 //
875 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
876 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
877 // CHECK2-NEXT:  entry:
878 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
879 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
880 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
881 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
882 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
883 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon.0*, align 8
884 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
885 // CHECK2-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
886 // CHECK2-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
887 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
888 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
889 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
890 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
891 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
892 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
893 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
894 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
895 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
896 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
897 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon.0*
898 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
899 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
900 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
901 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
902 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
903 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
904 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
905 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
906 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
907 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
908 // 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
909 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
910 // CHECK2-NEXT:    store %struct.anon.0* [[TMP8]], %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
911 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
912 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
913 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
914 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
915 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
916 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
917 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP12]], i32 0, i32 1
918 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
919 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
920 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
921 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
922 // CHECK2-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
923 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
924 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
925 // CHECK2-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
926 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
927 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 1
928 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
929 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
930 // CHECK2-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
931 // CHECK2-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
932 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
933 // CHECK2-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
934 // CHECK2-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
935 // CHECK2-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
936 // CHECK2-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
937 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
938 // CHECK2-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
939 // CHECK2-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
940 // CHECK2-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
941 // CHECK2-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
942 // CHECK2-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
943 // CHECK2-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
944 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
945 // CHECK2-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
946 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP12]], i32 0, i32 2
947 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
948 // CHECK2-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
949 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
950 // CHECK2-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
951 // CHECK2-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
952 // CHECK2-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
953 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
954 // CHECK2-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
955 // CHECK2-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
956 // CHECK2-NEXT:    ret i32 0
957 //
958 //
959 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
960 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
961 // CHECK2-NEXT:  entry:
962 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
963 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
964 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
965 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
966 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
967 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
968 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
969 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
970 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
971 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
972 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
973 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
974 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
975 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
976 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
977 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
978 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
979 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
980 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
981 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
982 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
983 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
984 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
985 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
986 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
987 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
988 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
989 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
990 // CHECK2:       omp.arraycpy.body:
991 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
992 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
993 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
994 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
995 // CHECK2-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
996 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
997 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
998 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
999 // CHECK2-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1000 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1001 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1002 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1003 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1004 // CHECK2:       omp.arraycpy.done5:
1005 // CHECK2-NEXT:    ret void
1006 //
1007