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 
main(int argc,char ** argv)13 int main(int argc, char **argv) {
14 #pragma omp parallel reduction(task, +: argc, argv[0:10][0:argc])
15   {
16 #pragma omp task in_reduction(+: argc, argv[0:10][0:argc])
17     ;
18   }
19 }
20 
21 
22 
23 // Init firstprivate copy of argc
24 
25 // Init firstprivate copy of argv[0:10][0:argc]
26 
27 // Register task reduction.
28 
29 
30 
31 
32 
33 
34 
35 
36 
37 #endif
38 // CHECK1-LABEL: define {{[^@]+}}@main
39 // CHECK1-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
40 // CHECK1-NEXT:  entry:
41 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
42 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
43 // CHECK1-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
44 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
45 // CHECK1-NEXT:    [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
46 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8** [[TMP0]])
47 // CHECK1-NEXT:    ret i32 0
48 //
49 //
50 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
51 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
52 // CHECK1-NEXT:  entry:
53 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
54 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
55 // CHECK1-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
56 // CHECK1-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
57 // CHECK1-NEXT:    [[ARGC1:%.*]] = alloca i32, align 4
58 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
59 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
60 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
61 // CHECK1-NEXT:    [[_TMP5:%.*]] = alloca i8*, align 8
62 // CHECK1-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
63 // CHECK1-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
64 // CHECK1-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
65 // CHECK1-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
66 // CHECK1-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
67 // CHECK1-NEXT:    [[_TMP24:%.*]] = alloca i8, align 1
68 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
69 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
70 // CHECK1-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
71 // CHECK1-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
72 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
73 // CHECK1-NEXT:    store i32 0, i32* [[ARGC1]], align 4
74 // CHECK1-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
75 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
76 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
77 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
78 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
79 // CHECK1-NEXT:    [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
80 // CHECK1-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
81 // CHECK1-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
82 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
83 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
84 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
85 // CHECK1-NEXT:    [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
86 // CHECK1-NEXT:    [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
87 // CHECK1-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
88 // CHECK1-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
89 // CHECK1-NEXT:    [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
90 // CHECK1-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
91 // CHECK1-NEXT:    [[TMP13:%.*]] = call i8* @llvm.stacksave()
92 // CHECK1-NEXT:    store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
93 // CHECK1-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
94 // CHECK1-NEXT:    store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
95 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
96 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
97 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
98 // CHECK1:       omp.arrayinit.body:
99 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
100 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
101 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
102 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
103 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
104 // CHECK1:       omp.arrayinit.done:
105 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
106 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
107 // CHECK1-NEXT:    [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
108 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
109 // CHECK1-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
110 // CHECK1-NEXT:    [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
111 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
112 // CHECK1-NEXT:    store i8** [[_TMP5]], i8*** [[TMP]], align 8
113 // CHECK1-NEXT:    store i8* [[TMP21]], i8** [[_TMP5]], align 8
114 // 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
115 // CHECK1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
116 // CHECK1-NEXT:    [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
117 // CHECK1-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
118 // CHECK1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
119 // CHECK1-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
120 // CHECK1-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
121 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
122 // CHECK1-NEXT:    store i64 4, i64* [[TMP26]], align 8
123 // CHECK1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
124 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
125 // CHECK1-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
126 // CHECK1-NEXT:    store i8* null, i8** [[TMP28]], align 8
127 // CHECK1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
128 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[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 6
130 // CHECK1-NEXT:    [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
131 // CHECK1-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
132 // CHECK1-NEXT:    [[DOTRD_INPUT_GEP_6:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
133 // CHECK1-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
134 // CHECK1-NEXT:    [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
135 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
136 // CHECK1-NEXT:    [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
137 // CHECK1-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
138 // CHECK1-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
139 // CHECK1-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
140 // CHECK1-NEXT:    [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
141 // CHECK1-NEXT:    [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
142 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
143 // CHECK1-NEXT:    [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
144 // CHECK1-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
145 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP32]], align 8
146 // CHECK1-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
147 // CHECK1-NEXT:    store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
148 // CHECK1-NEXT:    [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
149 // CHECK1-NEXT:    [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
150 // CHECK1-NEXT:    [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
151 // CHECK1-NEXT:    [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
152 // CHECK1-NEXT:    [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
153 // CHECK1-NEXT:    [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
154 // CHECK1-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
155 // CHECK1-NEXT:    store i64 [[TMP45]], i64* [[TMP46]], align 8
156 // CHECK1-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
157 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
158 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
159 // CHECK1-NEXT:    store i8* null, i8** [[TMP48]], align 8
160 // CHECK1-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
161 // CHECK1-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
162 // CHECK1-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
163 // CHECK1-NEXT:    store i32 1, i32* [[TMP50]], align 8
164 // CHECK1-NEXT:    [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
165 // CHECK1-NEXT:    [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
166 // CHECK1-NEXT:    [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
167 // CHECK1-NEXT:    [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 0, i32 2, i8* [[TMP53]])
168 // CHECK1-NEXT:    store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
169 // CHECK1-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
170 // CHECK1-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP55]], align 8
171 // CHECK1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
172 // CHECK1-NEXT:    store i32* [[ARGC1]], i32** [[TMP56]], align 8
173 // CHECK1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
174 // CHECK1-NEXT:    [[TMP58:%.*]] = load i8**, i8*** [[TMP]], align 8
175 // CHECK1-NEXT:    store i8** [[TMP58]], i8*** [[TMP57]], align 8
176 // CHECK1-NEXT:    [[TMP59:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
177 // CHECK1-NEXT:    [[TMP60:%.*]] = load i32, i32* [[TMP59]], align 4
178 // CHECK1-NEXT:    [[TMP61:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP60]], 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*)*))
179 // CHECK1-NEXT:    [[TMP62:%.*]] = bitcast i8* [[TMP61]] to %struct.kmp_task_t_with_privates*
180 // CHECK1-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP62]], i32 0, i32 0
181 // CHECK1-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP63]], i32 0, i32 0
182 // CHECK1-NEXT:    [[TMP65:%.*]] = load i8*, i8** [[TMP64]], align 8
183 // CHECK1-NEXT:    [[TMP66:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
184 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP65]], i8* align 8 [[TMP66]], i64 24, i1 false)
185 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP62]], i32 0, i32 1
186 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP67]], i32 0, i32 0
187 // CHECK1-NEXT:    [[TMP69:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
188 // CHECK1-NEXT:    store i8* [[TMP69]], i8** [[TMP68]], align 8
189 // CHECK1-NEXT:    [[TMP70:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
190 // CHECK1-NEXT:    [[TMP71:%.*]] = load i32, i32* [[TMP70]], align 4
191 // CHECK1-NEXT:    [[TMP72:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP71]], i8* [[TMP61]])
192 // CHECK1-NEXT:    [[TMP73:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
193 // CHECK1-NEXT:    [[TMP74:%.*]] = load i32, i32* [[TMP73]], align 4
194 // CHECK1-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP74]], i32 0)
195 // CHECK1-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
196 // CHECK1-NEXT:    [[TMP76:%.*]] = bitcast i32* [[ARGC1]] to i8*
197 // CHECK1-NEXT:    store i8* [[TMP76]], i8** [[TMP75]], align 8
198 // CHECK1-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
199 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP77]], align 8
200 // CHECK1-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
201 // CHECK1-NEXT:    [[TMP79:%.*]] = inttoptr i64 [[TMP11]] to i8*
202 // CHECK1-NEXT:    store i8* [[TMP79]], i8** [[TMP78]], align 8
203 // CHECK1-NEXT:    [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
204 // CHECK1-NEXT:    [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
205 // CHECK1-NEXT:    [[TMP82:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
206 // CHECK1-NEXT:    [[TMP83:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP81]], i32 2, i64 24, i8* [[TMP82]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
207 // CHECK1-NEXT:    switch i32 [[TMP83]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
208 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
209 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
210 // CHECK1-NEXT:    ]
211 // CHECK1:       .omp.reduction.case1:
212 // CHECK1-NEXT:    [[TMP84:%.*]] = load i32, i32* [[TMP0]], align 4
213 // CHECK1-NEXT:    [[TMP85:%.*]] = load i32, i32* [[ARGC1]], align 4
214 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP84]], [[TMP85]]
215 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP0]], align 4
216 // CHECK1-NEXT:    [[TMP86:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
217 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP86]]
218 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
219 // CHECK1:       omp.arraycpy.body:
220 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
221 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
222 // CHECK1-NEXT:    [[TMP87:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
223 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP87]] to i32
224 // CHECK1-NEXT:    [[TMP88:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
225 // CHECK1-NEXT:    [[CONV13:%.*]] = sext i8 [[TMP88]] to i32
226 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
227 // CHECK1-NEXT:    [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
228 // CHECK1-NEXT:    store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
229 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
230 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
231 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP86]]
232 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
233 // CHECK1:       omp.arraycpy.done18:
234 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP81]], [8 x i32]* @.gomp_critical_user_.reduction.var)
235 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
236 // CHECK1:       .omp.reduction.case2:
237 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, i32* [[ARGC1]], align 4
238 // CHECK1-NEXT:    [[TMP90:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP89]] monotonic, align 4
239 // CHECK1-NEXT:    [[TMP91:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
240 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP91]]
241 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
242 // CHECK1:       omp.arraycpy.body20:
243 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
244 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
245 // CHECK1-NEXT:    [[TMP92:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
246 // CHECK1-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP92]] to i32
247 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
248 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
249 // CHECK1:       atomic_cont:
250 // CHECK1-NEXT:    [[TMP93:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP98:%.*]], [[ATOMIC_CONT]] ]
251 // CHECK1-NEXT:    store i8 [[TMP93]], i8* [[_TMP24]], align 1
252 // CHECK1-NEXT:    [[TMP94:%.*]] = load i8, i8* [[_TMP24]], align 1
253 // CHECK1-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP94]] to i32
254 // CHECK1-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
255 // CHECK1-NEXT:    [[CONV26:%.*]] = sext i8 [[TMP95]] to i32
256 // CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
257 // CHECK1-NEXT:    [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
258 // CHECK1-NEXT:    store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
259 // CHECK1-NEXT:    [[TMP96:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
260 // CHECK1-NEXT:    [[TMP97:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP93]], i8 [[TMP96]] monotonic monotonic, align 1
261 // CHECK1-NEXT:    [[TMP98]] = extractvalue { i8, i1 } [[TMP97]], 0
262 // CHECK1-NEXT:    [[TMP99:%.*]] = extractvalue { i8, i1 } [[TMP97]], 1
263 // CHECK1-NEXT:    br i1 [[TMP99]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
264 // CHECK1:       atomic_exit:
265 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
266 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
267 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP91]]
268 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
269 // CHECK1:       omp.arraycpy.done32:
270 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
271 // CHECK1:       .omp.reduction.default:
272 // CHECK1-NEXT:    [[TMP100:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
273 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP100]])
274 // CHECK1-NEXT:    ret void
275 //
276 //
277 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
278 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
279 // CHECK1-NEXT:  entry:
280 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
281 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
282 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
283 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
284 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
285 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
286 // CHECK1-NEXT:    store i32 0, i32* [[TMP3]], align 8
287 // CHECK1-NEXT:    ret void
288 //
289 //
290 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
291 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
292 // CHECK1-NEXT:  entry:
293 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
294 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
295 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
296 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
297 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
298 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
299 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
300 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
301 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
302 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
303 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
304 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
305 // CHECK1-NEXT:    ret void
306 //
307 //
308 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
309 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
310 // CHECK1-NEXT:  entry:
311 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
312 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
313 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
314 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
315 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
316 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
317 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
318 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
319 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
320 // CHECK1:       omp.arrayinit.body:
321 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
322 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
323 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
324 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
325 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
326 // CHECK1:       omp.arrayinit.done:
327 // CHECK1-NEXT:    ret void
328 //
329 //
330 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
331 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
332 // CHECK1-NEXT:  entry:
333 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
334 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
335 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
336 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
337 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
338 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
339 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
340 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
341 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
342 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
343 // CHECK1:       omp.arraycpy.body:
344 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
345 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
346 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
347 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
348 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
349 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
350 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
351 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
352 // CHECK1-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
353 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
354 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
355 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
356 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
357 // CHECK1:       omp.arraycpy.done4:
358 // CHECK1-NEXT:    ret void
359 //
360 //
361 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
362 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
363 // CHECK1-NEXT:  entry:
364 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
365 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
366 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
367 // CHECK1-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
368 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
369 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
370 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
371 // CHECK1-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
372 // CHECK1-NEXT:    ret void
373 //
374 //
375 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
376 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
377 // CHECK1-NEXT:  entry:
378 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
379 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
380 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
381 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
382 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
383 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
384 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
385 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
386 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
387 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
389 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
390 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
391 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
392 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
393 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
394 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
395 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
396 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
397 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
398 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
399 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
400 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
401 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
402 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
403 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
404 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
405 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
406 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
407 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
408 // 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
409 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
410 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
411 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
412 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
413 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
414 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
415 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
416 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
417 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
418 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
419 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
420 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
421 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
422 // CHECK1-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
423 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
424 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
425 // CHECK1-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
426 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
427 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
428 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
429 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
430 // CHECK1-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
431 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
432 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
433 // CHECK1-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
434 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
435 // CHECK1-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
436 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
437 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
438 // CHECK1-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
439 // CHECK1-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
440 // CHECK1-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
441 // CHECK1-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
442 // CHECK1-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
443 // CHECK1-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
444 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
445 // CHECK1-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
446 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
447 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
448 // CHECK1-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
449 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
450 // CHECK1-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
451 // CHECK1-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
452 // CHECK1-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
453 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
454 // CHECK1-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
455 // CHECK1-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
456 // CHECK1-NEXT:    ret i32 0
457 //
458 //
459 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
460 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
461 // CHECK1-NEXT:  entry:
462 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
463 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
464 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
465 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
466 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
467 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
468 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
469 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
470 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
471 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
472 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
473 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
474 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
475 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
476 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
477 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
478 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
479 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
480 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
481 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
482 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
483 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
484 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
485 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
486 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
487 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
488 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
489 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
490 // CHECK1:       omp.arraycpy.body:
491 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
492 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
493 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
494 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
495 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
496 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
497 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
498 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
499 // CHECK1-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
500 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
501 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
502 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
503 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
504 // CHECK1:       omp.arraycpy.done5:
505 // CHECK1-NEXT:    ret void
506 //
507 //
508 // CHECK2-LABEL: define {{[^@]+}}@main
509 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
510 // CHECK2-NEXT:  entry:
511 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
512 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
513 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
514 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
515 // CHECK2-NEXT:    [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
516 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8** [[TMP0]])
517 // CHECK2-NEXT:    ret i32 0
518 //
519 //
520 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
521 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
522 // CHECK2-NEXT:  entry:
523 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
524 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
525 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
526 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
527 // CHECK2-NEXT:    [[ARGC1:%.*]] = alloca i32, align 4
528 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
529 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
530 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
531 // CHECK2-NEXT:    [[_TMP5:%.*]] = alloca i8*, align 8
532 // CHECK2-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
533 // CHECK2-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
534 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
535 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
536 // CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
537 // CHECK2-NEXT:    [[_TMP24:%.*]] = alloca i8, align 1
538 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
539 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
540 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
541 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
542 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
543 // CHECK2-NEXT:    store i32 0, i32* [[ARGC1]], align 4
544 // CHECK2-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
545 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
546 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
547 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
548 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
549 // CHECK2-NEXT:    [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
550 // CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
551 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
552 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
553 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
554 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
555 // CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
556 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
557 // CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
558 // CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
559 // CHECK2-NEXT:    [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
560 // CHECK2-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
561 // CHECK2-NEXT:    [[TMP13:%.*]] = call i8* @llvm.stacksave()
562 // CHECK2-NEXT:    store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
563 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
564 // CHECK2-NEXT:    store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
565 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
566 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
567 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
568 // CHECK2:       omp.arrayinit.body:
569 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
570 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
571 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
572 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
573 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
574 // CHECK2:       omp.arrayinit.done:
575 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
576 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
577 // CHECK2-NEXT:    [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
578 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
579 // CHECK2-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
580 // CHECK2-NEXT:    [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
581 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
582 // CHECK2-NEXT:    store i8** [[_TMP5]], i8*** [[TMP]], align 8
583 // CHECK2-NEXT:    store i8* [[TMP21]], i8** [[_TMP5]], align 8
584 // 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
585 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
586 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
587 // CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
588 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
589 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
590 // CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
591 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
592 // CHECK2-NEXT:    store i64 4, i64* [[TMP26]], align 8
593 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
594 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
595 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
596 // CHECK2-NEXT:    store i8* null, i8** [[TMP28]], align 8
597 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
598 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
599 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
600 // CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
601 // CHECK2-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
602 // CHECK2-NEXT:    [[DOTRD_INPUT_GEP_6:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
603 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
604 // CHECK2-NEXT:    [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
605 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
606 // CHECK2-NEXT:    [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
607 // CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
608 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
609 // CHECK2-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
610 // CHECK2-NEXT:    [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
611 // CHECK2-NEXT:    [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
612 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
613 // CHECK2-NEXT:    [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
614 // CHECK2-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
615 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP32]], align 8
616 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
617 // CHECK2-NEXT:    store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
618 // CHECK2-NEXT:    [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
619 // CHECK2-NEXT:    [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
620 // CHECK2-NEXT:    [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
621 // CHECK2-NEXT:    [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
622 // CHECK2-NEXT:    [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
623 // CHECK2-NEXT:    [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
624 // CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
625 // CHECK2-NEXT:    store i64 [[TMP45]], i64* [[TMP46]], align 8
626 // CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
627 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
628 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
629 // CHECK2-NEXT:    store i8* null, i8** [[TMP48]], align 8
630 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
631 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
632 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
633 // CHECK2-NEXT:    store i32 1, i32* [[TMP50]], align 8
634 // CHECK2-NEXT:    [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
635 // CHECK2-NEXT:    [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
636 // CHECK2-NEXT:    [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
637 // CHECK2-NEXT:    [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 0, i32 2, i8* [[TMP53]])
638 // CHECK2-NEXT:    store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
639 // CHECK2-NEXT:    [[TMP55:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
640 // CHECK2-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP55]], align 8
641 // CHECK2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
642 // CHECK2-NEXT:    store i32* [[ARGC1]], i32** [[TMP56]], align 8
643 // CHECK2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
644 // CHECK2-NEXT:    [[TMP58:%.*]] = load i8**, i8*** [[TMP]], align 8
645 // CHECK2-NEXT:    store i8** [[TMP58]], i8*** [[TMP57]], align 8
646 // CHECK2-NEXT:    [[TMP59:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
647 // CHECK2-NEXT:    [[TMP60:%.*]] = load i32, i32* [[TMP59]], align 4
648 // CHECK2-NEXT:    [[TMP61:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP60]], 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*)*))
649 // CHECK2-NEXT:    [[TMP62:%.*]] = bitcast i8* [[TMP61]] to %struct.kmp_task_t_with_privates*
650 // CHECK2-NEXT:    [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP62]], i32 0, i32 0
651 // CHECK2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP63]], i32 0, i32 0
652 // CHECK2-NEXT:    [[TMP65:%.*]] = load i8*, i8** [[TMP64]], align 8
653 // CHECK2-NEXT:    [[TMP66:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
654 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP65]], i8* align 8 [[TMP66]], i64 24, i1 false)
655 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP62]], i32 0, i32 1
656 // CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP67]], i32 0, i32 0
657 // CHECK2-NEXT:    [[TMP69:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
658 // CHECK2-NEXT:    store i8* [[TMP69]], i8** [[TMP68]], align 8
659 // CHECK2-NEXT:    [[TMP70:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
660 // CHECK2-NEXT:    [[TMP71:%.*]] = load i32, i32* [[TMP70]], align 4
661 // CHECK2-NEXT:    [[TMP72:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP71]], i8* [[TMP61]])
662 // CHECK2-NEXT:    [[TMP73:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
663 // CHECK2-NEXT:    [[TMP74:%.*]] = load i32, i32* [[TMP73]], align 4
664 // CHECK2-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP74]], i32 0)
665 // CHECK2-NEXT:    [[TMP75:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
666 // CHECK2-NEXT:    [[TMP76:%.*]] = bitcast i32* [[ARGC1]] to i8*
667 // CHECK2-NEXT:    store i8* [[TMP76]], i8** [[TMP75]], align 8
668 // CHECK2-NEXT:    [[TMP77:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
669 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP77]], align 8
670 // CHECK2-NEXT:    [[TMP78:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
671 // CHECK2-NEXT:    [[TMP79:%.*]] = inttoptr i64 [[TMP11]] to i8*
672 // CHECK2-NEXT:    store i8* [[TMP79]], i8** [[TMP78]], align 8
673 // CHECK2-NEXT:    [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
674 // CHECK2-NEXT:    [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
675 // CHECK2-NEXT:    [[TMP82:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
676 // CHECK2-NEXT:    [[TMP83:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP81]], i32 2, i64 24, i8* [[TMP82]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
677 // CHECK2-NEXT:    switch i32 [[TMP83]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
678 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
679 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
680 // CHECK2-NEXT:    ]
681 // CHECK2:       .omp.reduction.case1:
682 // CHECK2-NEXT:    [[TMP84:%.*]] = load i32, i32* [[TMP0]], align 4
683 // CHECK2-NEXT:    [[TMP85:%.*]] = load i32, i32* [[ARGC1]], align 4
684 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP84]], [[TMP85]]
685 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP0]], align 4
686 // CHECK2-NEXT:    [[TMP86:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
687 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP86]]
688 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
689 // CHECK2:       omp.arraycpy.body:
690 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
691 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
692 // CHECK2-NEXT:    [[TMP87:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
693 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP87]] to i32
694 // CHECK2-NEXT:    [[TMP88:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
695 // CHECK2-NEXT:    [[CONV13:%.*]] = sext i8 [[TMP88]] to i32
696 // CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
697 // CHECK2-NEXT:    [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
698 // CHECK2-NEXT:    store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
699 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
700 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
701 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP86]]
702 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
703 // CHECK2:       omp.arraycpy.done18:
704 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP81]], [8 x i32]* @.gomp_critical_user_.reduction.var)
705 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
706 // CHECK2:       .omp.reduction.case2:
707 // CHECK2-NEXT:    [[TMP89:%.*]] = load i32, i32* [[ARGC1]], align 4
708 // CHECK2-NEXT:    [[TMP90:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP89]] monotonic, align 4
709 // CHECK2-NEXT:    [[TMP91:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
710 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP91]]
711 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
712 // CHECK2:       omp.arraycpy.body20:
713 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
714 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
715 // CHECK2-NEXT:    [[TMP92:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
716 // CHECK2-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP92]] to i32
717 // CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
718 // CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
719 // CHECK2:       atomic_cont:
720 // CHECK2-NEXT:    [[TMP93:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP98:%.*]], [[ATOMIC_CONT]] ]
721 // CHECK2-NEXT:    store i8 [[TMP93]], i8* [[_TMP24]], align 1
722 // CHECK2-NEXT:    [[TMP94:%.*]] = load i8, i8* [[_TMP24]], align 1
723 // CHECK2-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP94]] to i32
724 // CHECK2-NEXT:    [[TMP95:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
725 // CHECK2-NEXT:    [[CONV26:%.*]] = sext i8 [[TMP95]] to i32
726 // CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
727 // CHECK2-NEXT:    [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
728 // CHECK2-NEXT:    store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
729 // CHECK2-NEXT:    [[TMP96:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
730 // CHECK2-NEXT:    [[TMP97:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP93]], i8 [[TMP96]] monotonic monotonic, align 1
731 // CHECK2-NEXT:    [[TMP98]] = extractvalue { i8, i1 } [[TMP97]], 0
732 // CHECK2-NEXT:    [[TMP99:%.*]] = extractvalue { i8, i1 } [[TMP97]], 1
733 // CHECK2-NEXT:    br i1 [[TMP99]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
734 // CHECK2:       atomic_exit:
735 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
736 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
737 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP91]]
738 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
739 // CHECK2:       omp.arraycpy.done32:
740 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
741 // CHECK2:       .omp.reduction.default:
742 // CHECK2-NEXT:    [[TMP100:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
743 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP100]])
744 // CHECK2-NEXT:    ret void
745 //
746 //
747 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
748 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
749 // CHECK2-NEXT:  entry:
750 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
751 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
752 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
753 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
754 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
755 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
756 // CHECK2-NEXT:    store i32 0, i32* [[TMP3]], align 8
757 // CHECK2-NEXT:    ret void
758 //
759 //
760 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
761 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
762 // CHECK2-NEXT:  entry:
763 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
764 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
765 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
766 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
767 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
768 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
769 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
770 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
771 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
772 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
773 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
774 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
775 // CHECK2-NEXT:    ret void
776 //
777 //
778 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
779 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
780 // CHECK2-NEXT:  entry:
781 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
782 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
783 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
784 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
785 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
786 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
787 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
788 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
789 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
790 // CHECK2:       omp.arrayinit.body:
791 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
792 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
793 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
794 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
795 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
796 // CHECK2:       omp.arrayinit.done:
797 // CHECK2-NEXT:    ret void
798 //
799 //
800 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
801 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
802 // CHECK2-NEXT:  entry:
803 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
804 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
805 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
806 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
807 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
808 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
809 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
810 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
811 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
812 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
813 // CHECK2:       omp.arraycpy.body:
814 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
815 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
816 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
817 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
818 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
819 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
820 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
821 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
822 // CHECK2-NEXT:    store i8 [[CONV3]], 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_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
825 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
826 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
827 // CHECK2:       omp.arraycpy.done4:
828 // CHECK2-NEXT:    ret void
829 //
830 //
831 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
832 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
833 // CHECK2-NEXT:  entry:
834 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
835 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
836 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
837 // CHECK2-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
838 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
839 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
840 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
841 // CHECK2-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
842 // CHECK2-NEXT:    ret void
843 //
844 //
845 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
846 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
847 // CHECK2-NEXT:  entry:
848 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
849 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
850 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
851 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
852 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
853 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
854 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
855 // CHECK2-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
856 // CHECK2-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
857 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
858 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
859 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
860 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
861 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
862 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
863 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
864 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
865 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
866 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
867 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
868 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
869 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
870 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
871 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
872 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
873 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
874 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
875 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
876 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
877 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
878 // 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
879 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
880 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
881 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
882 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
883 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
884 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
885 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
886 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
887 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
888 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
889 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
890 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
891 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
892 // CHECK2-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
893 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
894 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
895 // CHECK2-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
896 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
897 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
898 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
899 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
900 // CHECK2-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
901 // CHECK2-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
902 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
903 // CHECK2-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
904 // CHECK2-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
905 // CHECK2-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
906 // CHECK2-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
907 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
908 // CHECK2-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
909 // CHECK2-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
910 // CHECK2-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
911 // CHECK2-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
912 // CHECK2-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
913 // CHECK2-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
914 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
915 // CHECK2-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
916 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
917 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
918 // CHECK2-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
919 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
920 // CHECK2-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
921 // CHECK2-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
922 // CHECK2-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
923 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
924 // CHECK2-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
925 // CHECK2-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
926 // CHECK2-NEXT:    ret i32 0
927 //
928 //
929 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
930 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
931 // CHECK2-NEXT:  entry:
932 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
933 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
934 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
935 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
936 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
937 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
938 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
939 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
940 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
941 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
942 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
943 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
944 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
945 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
946 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
947 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
948 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
949 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
950 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
951 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
952 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
953 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
954 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
955 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
956 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
957 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
958 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
959 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
960 // CHECK2:       omp.arraycpy.body:
961 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
962 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
963 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
964 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
965 // CHECK2-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
966 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
967 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
968 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
969 // CHECK2-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
970 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
971 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
972 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
973 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
974 // CHECK2:       omp.arraycpy.done5:
975 // CHECK2-NEXT:    ret void
976 //
977