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 noundef [[ARGC:%.*]], i8** noundef [[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 noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** noundef [[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:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
170 // CHECK1-NEXT:    [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
171 // CHECK1-NEXT:    [[TMP57:%.*]] = call i32 @__kmpc_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP56]])
172 // CHECK1-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
173 // CHECK1-NEXT:    br i1 [[TMP58]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_END:%.*]]
174 // CHECK1:       omp_if.then:
175 // CHECK1-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
176 // CHECK1-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP59]], align 8
177 // CHECK1-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
178 // CHECK1-NEXT:    store i32* [[ARGC1]], i32** [[TMP60]], align 8
179 // CHECK1-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
180 // CHECK1-NEXT:    [[TMP62:%.*]] = load i8**, i8*** [[TMP]], align 8
181 // CHECK1-NEXT:    store i8** [[TMP62]], i8*** [[TMP61]], align 8
182 // CHECK1-NEXT:    [[TMP63:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
183 // CHECK1-NEXT:    [[TMP64:%.*]] = load i32, i32* [[TMP63]], align 4
184 // CHECK1-NEXT:    [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP64]], i32 1, i64 48, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @.omp_task_entry. to i32 (i32, i8*)*))
185 // CHECK1-NEXT:    [[TMP66:%.*]] = bitcast i8* [[TMP65]] to %struct.kmp_task_t_with_privates*
186 // CHECK1-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP66]], i32 0, i32 0
187 // CHECK1-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP67]], i32 0, i32 0
188 // CHECK1-NEXT:    [[TMP69:%.*]] = load i8*, i8** [[TMP68]], align 8
189 // CHECK1-NEXT:    [[TMP70:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
190 // CHECK1-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP69]], i8* align 8 [[TMP70]], i64 24, i1 false)
191 // CHECK1-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP66]], i32 0, i32 1
192 // CHECK1-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP71]], i32 0, i32 0
193 // CHECK1-NEXT:    [[TMP73:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
194 // CHECK1-NEXT:    store i8* [[TMP73]], i8** [[TMP72]], align 8
195 // CHECK1-NEXT:    [[TMP74:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
196 // CHECK1-NEXT:    [[TMP75:%.*]] = load i32, i32* [[TMP74]], align 4
197 // CHECK1-NEXT:    [[TMP76:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP75]], i8* [[TMP65]])
198 // CHECK1-NEXT:    call void @__kmpc_end_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP56]])
199 // CHECK1-NEXT:    br label [[OMP_IF_END]]
200 // CHECK1:       omp_if.end:
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:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP78]], i32 0)
204 // CHECK1-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
205 // CHECK1-NEXT:    [[TMP80:%.*]] = bitcast i32* [[ARGC1]] to i8*
206 // CHECK1-NEXT:    store i8* [[TMP80]], i8** [[TMP79]], align 8
207 // CHECK1-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
208 // CHECK1-NEXT:    store i8* [[VLA]], i8** [[TMP81]], align 8
209 // CHECK1-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
210 // CHECK1-NEXT:    [[TMP83:%.*]] = inttoptr i64 [[TMP11]] to i8*
211 // CHECK1-NEXT:    store i8* [[TMP83]], i8** [[TMP82]], align 8
212 // CHECK1-NEXT:    [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
213 // CHECK1-NEXT:    [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
214 // CHECK1-NEXT:    [[TMP86:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
215 // CHECK1-NEXT:    [[TMP87:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP85]], i32 2, i64 24, i8* [[TMP86]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
216 // CHECK1-NEXT:    switch i32 [[TMP87]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
217 // CHECK1-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
218 // CHECK1-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
219 // CHECK1-NEXT:    ]
220 // CHECK1:       .omp.reduction.case1:
221 // CHECK1-NEXT:    [[TMP88:%.*]] = load i32, i32* [[TMP0]], align 4
222 // CHECK1-NEXT:    [[TMP89:%.*]] = load i32, i32* [[ARGC1]], align 4
223 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP88]], [[TMP89]]
224 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP0]], align 4
225 // CHECK1-NEXT:    [[TMP90:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
226 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP90]]
227 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
228 // CHECK1:       omp.arraycpy.body:
229 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
230 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
231 // CHECK1-NEXT:    [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
232 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP91]] to i32
233 // CHECK1-NEXT:    [[TMP92:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
234 // CHECK1-NEXT:    [[CONV13:%.*]] = sext i8 [[TMP92]] to i32
235 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
236 // CHECK1-NEXT:    [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
237 // CHECK1-NEXT:    store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
238 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
239 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
240 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP90]]
241 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
242 // CHECK1:       omp.arraycpy.done18:
243 // CHECK1-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]], [8 x i32]* @.gomp_critical_user_.reduction.var)
244 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
245 // CHECK1:       .omp.reduction.case2:
246 // CHECK1-NEXT:    [[TMP93:%.*]] = load i32, i32* [[ARGC1]], align 4
247 // CHECK1-NEXT:    [[TMP94:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP93]] monotonic, align 4
248 // CHECK1-NEXT:    [[TMP95:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
249 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP95]]
250 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
251 // CHECK1:       omp.arraycpy.body20:
252 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
253 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
254 // CHECK1-NEXT:    [[TMP96:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
255 // CHECK1-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP96]] to i32
256 // CHECK1-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
257 // CHECK1-NEXT:    br label [[ATOMIC_CONT:%.*]]
258 // CHECK1:       atomic_cont:
259 // CHECK1-NEXT:    [[TMP97:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP102:%.*]], [[ATOMIC_CONT]] ]
260 // CHECK1-NEXT:    store i8 [[TMP97]], i8* [[_TMP24]], align 1
261 // CHECK1-NEXT:    [[TMP98:%.*]] = load i8, i8* [[_TMP24]], align 1
262 // CHECK1-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP98]] to i32
263 // CHECK1-NEXT:    [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
264 // CHECK1-NEXT:    [[CONV26:%.*]] = sext i8 [[TMP99]] to i32
265 // CHECK1-NEXT:    [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
266 // CHECK1-NEXT:    [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
267 // CHECK1-NEXT:    store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
268 // CHECK1-NEXT:    [[TMP100:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
269 // CHECK1-NEXT:    [[TMP101:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP97]], i8 [[TMP100]] monotonic monotonic, align 1
270 // CHECK1-NEXT:    [[TMP102]] = extractvalue { i8, i1 } [[TMP101]], 0
271 // CHECK1-NEXT:    [[TMP103:%.*]] = extractvalue { i8, i1 } [[TMP101]], 1
272 // CHECK1-NEXT:    br i1 [[TMP103]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
273 // CHECK1:       atomic_exit:
274 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
275 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
276 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP95]]
277 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
278 // CHECK1:       omp.arraycpy.done32:
279 // CHECK1-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
280 // CHECK1:       .omp.reduction.default:
281 // CHECK1-NEXT:    [[TMP104:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
282 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP104]])
283 // CHECK1-NEXT:    ret void
284 //
285 //
286 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
287 // CHECK1-SAME: (i8* noalias noundef [[TMP0:%.*]], i8* noalias noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
288 // CHECK1-NEXT:  entry:
289 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
290 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
291 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
292 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
293 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
294 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
295 // CHECK1-NEXT:    store i32 0, i32* [[TMP3]], align 8
296 // CHECK1-NEXT:    ret void
297 //
298 //
299 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
300 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
301 // CHECK1-NEXT:  entry:
302 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
303 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
304 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
305 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
306 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
307 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
308 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
309 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
310 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
311 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
312 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
313 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
314 // CHECK1-NEXT:    ret void
315 //
316 //
317 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
318 // CHECK1-SAME: (i8* noalias noundef [[TMP0:%.*]], i8* noalias noundef [[TMP1:%.*]]) #[[ATTR3]] {
319 // CHECK1-NEXT:  entry:
320 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
321 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
322 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
323 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
324 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
325 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
326 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
327 // CHECK1-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
328 // CHECK1-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
329 // CHECK1:       omp.arrayinit.body:
330 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
331 // CHECK1-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
332 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
333 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
334 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
335 // CHECK1:       omp.arrayinit.done:
336 // CHECK1-NEXT:    ret void
337 //
338 //
339 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
340 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
341 // CHECK1-NEXT:  entry:
342 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
343 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
344 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
345 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
346 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
347 // CHECK1-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
348 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
349 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
350 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
351 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
352 // CHECK1:       omp.arraycpy.body:
353 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
354 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
355 // CHECK1-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
356 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
357 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
358 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
359 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
360 // CHECK1-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
361 // CHECK1-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
362 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
363 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
364 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
365 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
366 // CHECK1:       omp.arraycpy.done4:
367 // CHECK1-NEXT:    ret void
368 //
369 //
370 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
371 // CHECK1-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i8*** noalias noundef [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
372 // CHECK1-NEXT:  entry:
373 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
374 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
375 // CHECK1-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
376 // CHECK1-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
377 // CHECK1-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
378 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
379 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
380 // CHECK1-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
381 // CHECK1-NEXT:    ret void
382 //
383 //
384 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
385 // CHECK1-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR3]] {
386 // CHECK1-NEXT:  entry:
387 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
389 // CHECK1-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
390 // CHECK1-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
391 // CHECK1-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
392 // CHECK1-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
393 // CHECK1-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
394 // CHECK1-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
395 // CHECK1-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
396 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
397 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
398 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
399 // CHECK1-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
400 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
401 // CHECK1-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
402 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
403 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
404 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
405 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
406 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
407 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
408 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
409 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
410 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
411 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
412 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
413 // CHECK1-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
414 // CHECK1-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
415 // CHECK1-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
416 // CHECK1-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
417 // 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
418 // CHECK1-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
419 // CHECK1-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
420 // CHECK1-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
421 // CHECK1-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
422 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
423 // CHECK1-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
424 // CHECK1-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
425 // CHECK1-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
426 // CHECK1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
427 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
428 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
429 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
430 // CHECK1-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
431 // CHECK1-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
432 // CHECK1-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
433 // CHECK1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
434 // CHECK1-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
435 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
436 // CHECK1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
437 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
438 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
439 // CHECK1-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
440 // CHECK1-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
441 // CHECK1-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
442 // CHECK1-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
443 // CHECK1-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
444 // CHECK1-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
445 // CHECK1-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
446 // CHECK1-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
447 // CHECK1-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
448 // CHECK1-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
449 // CHECK1-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
450 // CHECK1-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
451 // CHECK1-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
452 // CHECK1-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
453 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
454 // CHECK1-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
455 // CHECK1-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
456 // CHECK1-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
457 // CHECK1-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
458 // CHECK1-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
459 // CHECK1-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
460 // CHECK1-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
461 // CHECK1-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
462 // CHECK1-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
463 // CHECK1-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
464 // CHECK1-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
465 // CHECK1-NEXT:    ret i32 0
466 //
467 //
468 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
469 // CHECK1-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
470 // CHECK1-NEXT:  entry:
471 // CHECK1-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
472 // CHECK1-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
473 // CHECK1-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
474 // CHECK1-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
475 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
476 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
477 // CHECK1-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
478 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
479 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
480 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
481 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
482 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
483 // CHECK1-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
484 // CHECK1-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
485 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
486 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
487 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
488 // CHECK1-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
489 // CHECK1-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
490 // CHECK1-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
491 // CHECK1-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
492 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
493 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
494 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
495 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
496 // CHECK1-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
497 // CHECK1-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
498 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
499 // CHECK1:       omp.arraycpy.body:
500 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
501 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
502 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
503 // CHECK1-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
504 // CHECK1-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
505 // CHECK1-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
506 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
507 // CHECK1-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
508 // CHECK1-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
509 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
510 // CHECK1-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
511 // CHECK1-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
512 // CHECK1-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
513 // CHECK1:       omp.arraycpy.done5:
514 // CHECK1-NEXT:    ret void
515 //
516 //
517 // CHECK2-LABEL: define {{[^@]+}}@main
518 // CHECK2-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
519 // CHECK2-NEXT:  entry:
520 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32, align 4
521 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
522 // CHECK2-NEXT:    store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
523 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
524 // CHECK2-NEXT:    [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
525 // 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]])
526 // CHECK2-NEXT:    ret i32 0
527 //
528 //
529 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
530 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
531 // CHECK2-NEXT:  entry:
532 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
533 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
534 // CHECK2-NEXT:    [[ARGC_ADDR:%.*]] = alloca i32*, align 8
535 // CHECK2-NEXT:    [[ARGV_ADDR:%.*]] = alloca i8**, align 8
536 // CHECK2-NEXT:    [[ARGC1:%.*]] = alloca i32, align 4
537 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
538 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
539 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8**, align 8
540 // CHECK2-NEXT:    [[_TMP5:%.*]] = alloca i8*, align 8
541 // CHECK2-NEXT:    [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
542 // CHECK2-NEXT:    [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
543 // CHECK2-NEXT:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
544 // CHECK2-NEXT:    [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
545 // CHECK2-NEXT:    [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
546 // CHECK2-NEXT:    [[_TMP24:%.*]] = alloca i8, align 1
547 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
548 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
549 // CHECK2-NEXT:    store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
550 // CHECK2-NEXT:    store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
551 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
552 // CHECK2-NEXT:    store i32 0, i32* [[ARGC1]], align 4
553 // CHECK2-NEXT:    [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
554 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
555 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
556 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
557 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
558 // CHECK2-NEXT:    [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
559 // CHECK2-NEXT:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
560 // CHECK2-NEXT:    [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
561 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
562 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
563 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
564 // CHECK2-NEXT:    [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
565 // CHECK2-NEXT:    [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
566 // CHECK2-NEXT:    [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
567 // CHECK2-NEXT:    [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
568 // CHECK2-NEXT:    [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
569 // CHECK2-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
570 // CHECK2-NEXT:    [[TMP13:%.*]] = call i8* @llvm.stacksave()
571 // CHECK2-NEXT:    store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
572 // CHECK2-NEXT:    [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
573 // CHECK2-NEXT:    store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
574 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
575 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
576 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
577 // CHECK2:       omp.arrayinit.body:
578 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
579 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
580 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
581 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
582 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
583 // CHECK2:       omp.arrayinit.done:
584 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
585 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
586 // CHECK2-NEXT:    [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
587 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
588 // CHECK2-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
589 // CHECK2-NEXT:    [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
590 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
591 // CHECK2-NEXT:    store i8** [[_TMP5]], i8*** [[TMP]], align 8
592 // CHECK2-NEXT:    store i8* [[TMP21]], i8** [[_TMP5]], align 8
593 // 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
594 // CHECK2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
595 // CHECK2-NEXT:    [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
596 // CHECK2-NEXT:    store i8* [[TMP23]], i8** [[TMP22]], align 8
597 // CHECK2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
598 // CHECK2-NEXT:    [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
599 // CHECK2-NEXT:    store i8* [[TMP25]], i8** [[TMP24]], align 8
600 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
601 // CHECK2-NEXT:    store i64 4, i64* [[TMP26]], align 8
602 // CHECK2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
603 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
604 // CHECK2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
605 // CHECK2-NEXT:    store i8* null, i8** [[TMP28]], align 8
606 // CHECK2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
607 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
608 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
609 // CHECK2-NEXT:    [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
610 // CHECK2-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
611 // 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
612 // CHECK2-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
613 // CHECK2-NEXT:    [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
614 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
615 // CHECK2-NEXT:    [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
616 // CHECK2-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
617 // CHECK2-NEXT:    [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
618 // CHECK2-NEXT:    [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
619 // CHECK2-NEXT:    [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
620 // CHECK2-NEXT:    [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
621 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
622 // CHECK2-NEXT:    [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
623 // CHECK2-NEXT:    [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
624 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP32]], align 8
625 // CHECK2-NEXT:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
626 // CHECK2-NEXT:    store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
627 // CHECK2-NEXT:    [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
628 // CHECK2-NEXT:    [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
629 // CHECK2-NEXT:    [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
630 // CHECK2-NEXT:    [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
631 // CHECK2-NEXT:    [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
632 // CHECK2-NEXT:    [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
633 // CHECK2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
634 // CHECK2-NEXT:    store i64 [[TMP45]], i64* [[TMP46]], align 8
635 // CHECK2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
636 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
637 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
638 // CHECK2-NEXT:    store i8* null, i8** [[TMP48]], align 8
639 // CHECK2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
640 // CHECK2-NEXT:    store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
641 // CHECK2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
642 // CHECK2-NEXT:    store i32 1, i32* [[TMP50]], align 8
643 // CHECK2-NEXT:    [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
644 // CHECK2-NEXT:    [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
645 // CHECK2-NEXT:    [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
646 // CHECK2-NEXT:    [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 0, i32 2, i8* [[TMP53]])
647 // CHECK2-NEXT:    store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
648 // CHECK2-NEXT:    [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
649 // CHECK2-NEXT:    [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
650 // CHECK2-NEXT:    [[TMP57:%.*]] = call i32 @__kmpc_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP56]])
651 // CHECK2-NEXT:    [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
652 // CHECK2-NEXT:    br i1 [[TMP58]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_END:%.*]]
653 // CHECK2:       omp_if.then:
654 // CHECK2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
655 // CHECK2-NEXT:    store i8** [[DOTTASK_RED_]], i8*** [[TMP59]], align 8
656 // CHECK2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
657 // CHECK2-NEXT:    store i32* [[ARGC1]], i32** [[TMP60]], align 8
658 // CHECK2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
659 // CHECK2-NEXT:    [[TMP62:%.*]] = load i8**, i8*** [[TMP]], align 8
660 // CHECK2-NEXT:    store i8** [[TMP62]], i8*** [[TMP61]], align 8
661 // CHECK2-NEXT:    [[TMP63:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
662 // CHECK2-NEXT:    [[TMP64:%.*]] = load i32, i32* [[TMP63]], align 4
663 // CHECK2-NEXT:    [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP64]], 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*)*))
664 // CHECK2-NEXT:    [[TMP66:%.*]] = bitcast i8* [[TMP65]] to %struct.kmp_task_t_with_privates*
665 // CHECK2-NEXT:    [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP66]], i32 0, i32 0
666 // CHECK2-NEXT:    [[TMP68:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP67]], i32 0, i32 0
667 // CHECK2-NEXT:    [[TMP69:%.*]] = load i8*, i8** [[TMP68]], align 8
668 // CHECK2-NEXT:    [[TMP70:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
669 // CHECK2-NEXT:    call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP69]], i8* align 8 [[TMP70]], i64 24, i1 false)
670 // CHECK2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP66]], i32 0, i32 1
671 // CHECK2-NEXT:    [[TMP72:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP71]], i32 0, i32 0
672 // CHECK2-NEXT:    [[TMP73:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
673 // CHECK2-NEXT:    store i8* [[TMP73]], i8** [[TMP72]], align 8
674 // CHECK2-NEXT:    [[TMP74:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
675 // CHECK2-NEXT:    [[TMP75:%.*]] = load i32, i32* [[TMP74]], align 4
676 // CHECK2-NEXT:    [[TMP76:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP75]], i8* [[TMP65]])
677 // CHECK2-NEXT:    call void @__kmpc_end_master(%struct.ident_t* @[[GLOB1]], i32 [[TMP56]])
678 // CHECK2-NEXT:    br label [[OMP_IF_END]]
679 // CHECK2:       omp_if.end:
680 // CHECK2-NEXT:    [[TMP77:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
681 // CHECK2-NEXT:    [[TMP78:%.*]] = load i32, i32* [[TMP77]], align 4
682 // CHECK2-NEXT:    call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP78]], i32 0)
683 // CHECK2-NEXT:    [[TMP79:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
684 // CHECK2-NEXT:    [[TMP80:%.*]] = bitcast i32* [[ARGC1]] to i8*
685 // CHECK2-NEXT:    store i8* [[TMP80]], i8** [[TMP79]], align 8
686 // CHECK2-NEXT:    [[TMP81:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
687 // CHECK2-NEXT:    store i8* [[VLA]], i8** [[TMP81]], align 8
688 // CHECK2-NEXT:    [[TMP82:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
689 // CHECK2-NEXT:    [[TMP83:%.*]] = inttoptr i64 [[TMP11]] to i8*
690 // CHECK2-NEXT:    store i8* [[TMP83]], i8** [[TMP82]], align 8
691 // CHECK2-NEXT:    [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
692 // CHECK2-NEXT:    [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
693 // CHECK2-NEXT:    [[TMP86:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
694 // CHECK2-NEXT:    [[TMP87:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP85]], i32 2, i64 24, i8* [[TMP86]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
695 // CHECK2-NEXT:    switch i32 [[TMP87]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
696 // CHECK2-NEXT:    i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
697 // CHECK2-NEXT:    i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
698 // CHECK2-NEXT:    ]
699 // CHECK2:       .omp.reduction.case1:
700 // CHECK2-NEXT:    [[TMP88:%.*]] = load i32, i32* [[TMP0]], align 4
701 // CHECK2-NEXT:    [[TMP89:%.*]] = load i32, i32* [[ARGC1]], align 4
702 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP88]], [[TMP89]]
703 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP0]], align 4
704 // CHECK2-NEXT:    [[TMP90:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
705 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP90]]
706 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
707 // CHECK2:       omp.arraycpy.body:
708 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
709 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
710 // CHECK2-NEXT:    [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
711 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP91]] to i32
712 // CHECK2-NEXT:    [[TMP92:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
713 // CHECK2-NEXT:    [[CONV13:%.*]] = sext i8 [[TMP92]] to i32
714 // CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
715 // CHECK2-NEXT:    [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
716 // CHECK2-NEXT:    store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
717 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
718 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
719 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP90]]
720 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
721 // CHECK2:       omp.arraycpy.done18:
722 // CHECK2-NEXT:    call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]], [8 x i32]* @.gomp_critical_user_.reduction.var)
723 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
724 // CHECK2:       .omp.reduction.case2:
725 // CHECK2-NEXT:    [[TMP93:%.*]] = load i32, i32* [[ARGC1]], align 4
726 // CHECK2-NEXT:    [[TMP94:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP93]] monotonic, align 4
727 // CHECK2-NEXT:    [[TMP95:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
728 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP95]]
729 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
730 // CHECK2:       omp.arraycpy.body20:
731 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
732 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
733 // CHECK2-NEXT:    [[TMP96:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
734 // CHECK2-NEXT:    [[CONV23:%.*]] = sext i8 [[TMP96]] to i32
735 // CHECK2-NEXT:    [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
736 // CHECK2-NEXT:    br label [[ATOMIC_CONT:%.*]]
737 // CHECK2:       atomic_cont:
738 // CHECK2-NEXT:    [[TMP97:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP102:%.*]], [[ATOMIC_CONT]] ]
739 // CHECK2-NEXT:    store i8 [[TMP97]], i8* [[_TMP24]], align 1
740 // CHECK2-NEXT:    [[TMP98:%.*]] = load i8, i8* [[_TMP24]], align 1
741 // CHECK2-NEXT:    [[CONV25:%.*]] = sext i8 [[TMP98]] to i32
742 // CHECK2-NEXT:    [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
743 // CHECK2-NEXT:    [[CONV26:%.*]] = sext i8 [[TMP99]] to i32
744 // CHECK2-NEXT:    [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
745 // CHECK2-NEXT:    [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
746 // CHECK2-NEXT:    store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
747 // CHECK2-NEXT:    [[TMP100:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
748 // CHECK2-NEXT:    [[TMP101:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP97]], i8 [[TMP100]] monotonic monotonic, align 1
749 // CHECK2-NEXT:    [[TMP102]] = extractvalue { i8, i1 } [[TMP101]], 0
750 // CHECK2-NEXT:    [[TMP103:%.*]] = extractvalue { i8, i1 } [[TMP101]], 1
751 // CHECK2-NEXT:    br i1 [[TMP103]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
752 // CHECK2:       atomic_exit:
753 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
754 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
755 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP95]]
756 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
757 // CHECK2:       omp.arraycpy.done32:
758 // CHECK2-NEXT:    br label [[DOTOMP_REDUCTION_DEFAULT]]
759 // CHECK2:       .omp.reduction.default:
760 // CHECK2-NEXT:    [[TMP104:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
761 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP104]])
762 // CHECK2-NEXT:    ret void
763 //
764 //
765 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
766 // CHECK2-SAME: (i8* noalias noundef [[TMP0:%.*]], i8* noalias noundef [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
767 // CHECK2-NEXT:  entry:
768 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
769 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
770 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
771 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
772 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
773 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
774 // CHECK2-NEXT:    store i32 0, i32* [[TMP3]], align 8
775 // CHECK2-NEXT:    ret void
776 //
777 //
778 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
779 // CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[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:%.*]] = bitcast i8* [[TMP2]] to i32*
787 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
788 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
789 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
790 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
791 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
792 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP3]], align 8
793 // CHECK2-NEXT:    ret void
794 //
795 //
796 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
797 // CHECK2-SAME: (i8* noalias noundef [[TMP0:%.*]], i8* noalias noundef [[TMP1:%.*]]) #[[ATTR3]] {
798 // CHECK2-NEXT:  entry:
799 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
800 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
801 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
802 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
803 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
804 // CHECK2-NEXT:    [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
805 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
806 // CHECK2-NEXT:    [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
807 // CHECK2-NEXT:    br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
808 // CHECK2:       omp.arrayinit.body:
809 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
810 // CHECK2-NEXT:    store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
811 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
812 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
813 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
814 // CHECK2:       omp.arrayinit.done:
815 // CHECK2-NEXT:    ret void
816 //
817 //
818 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
819 // CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
820 // CHECK2-NEXT:  entry:
821 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
822 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
823 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
824 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
825 // CHECK2-NEXT:    [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
826 // CHECK2-NEXT:    [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
827 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
828 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
829 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
830 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
831 // CHECK2:       omp.arraycpy.body:
832 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
833 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
834 // CHECK2-NEXT:    [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
835 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP6]] to i32
836 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
837 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
838 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
839 // CHECK2-NEXT:    [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
840 // CHECK2-NEXT:    store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
841 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
842 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
843 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
844 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
845 // CHECK2:       omp.arraycpy.done4:
846 // CHECK2-NEXT:    ret void
847 //
848 //
849 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
850 // CHECK2-SAME: (%struct..kmp_privates.t* noalias noundef [[TMP0:%.*]], i8*** noalias noundef [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
851 // CHECK2-NEXT:  entry:
852 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
853 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8***, align 8
854 // CHECK2-NEXT:    store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
855 // CHECK2-NEXT:    store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
856 // CHECK2-NEXT:    [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
857 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
858 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
859 // CHECK2-NEXT:    store i8** [[TMP3]], i8*** [[TMP4]], align 8
860 // CHECK2-NEXT:    ret void
861 //
862 //
863 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
864 // CHECK2-SAME: (i32 noundef [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias noundef [[TMP1:%.*]]) #[[ATTR3]] {
865 // CHECK2-NEXT:  entry:
866 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
867 // CHECK2-NEXT:    [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
868 // CHECK2-NEXT:    [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
869 // CHECK2-NEXT:    [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
870 // CHECK2-NEXT:    [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
871 // CHECK2-NEXT:    [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
872 // CHECK2-NEXT:    [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
873 // CHECK2-NEXT:    [[TMP_I:%.*]] = alloca i8**, align 8
874 // CHECK2-NEXT:    [[TMP4_I:%.*]] = alloca i8*, align 8
875 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i32, align 4
876 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
877 // CHECK2-NEXT:    store i32 [[TMP0]], i32* [[DOTADDR]], align 4
878 // CHECK2-NEXT:    store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
879 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
880 // CHECK2-NEXT:    [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
881 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
882 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
883 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
884 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
885 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
886 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
887 // CHECK2-NEXT:    [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
888 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
889 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
890 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
891 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
892 // CHECK2-NEXT:    call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
893 // CHECK2-NEXT:    store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
894 // CHECK2-NEXT:    store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
895 // CHECK2-NEXT:    store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
896 // 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
897 // CHECK2-NEXT:    store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
898 // CHECK2-NEXT:    store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
899 // CHECK2-NEXT:    [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
900 // CHECK2-NEXT:    [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
901 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
902 // CHECK2-NEXT:    [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
903 // CHECK2-NEXT:    call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
904 // CHECK2-NEXT:    [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
905 // CHECK2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
906 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
907 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
908 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
909 // CHECK2-NEXT:    [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
910 // CHECK2-NEXT:    [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
911 // CHECK2-NEXT:    [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
912 // CHECK2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
913 // CHECK2-NEXT:    [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
914 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
915 // CHECK2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
916 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
917 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
918 // CHECK2-NEXT:    [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
919 // CHECK2-NEXT:    [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
920 // CHECK2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
921 // CHECK2-NEXT:    [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
922 // CHECK2-NEXT:    [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
923 // CHECK2-NEXT:    [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
924 // CHECK2-NEXT:    [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
925 // CHECK2-NEXT:    [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
926 // CHECK2-NEXT:    [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
927 // CHECK2-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
928 // CHECK2-NEXT:    [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
929 // CHECK2-NEXT:    [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
930 // CHECK2-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
931 // CHECK2-NEXT:    store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
932 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
933 // CHECK2-NEXT:    [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
934 // CHECK2-NEXT:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
935 // CHECK2-NEXT:    [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
936 // CHECK2-NEXT:    [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
937 // CHECK2-NEXT:    [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
938 // CHECK2-NEXT:    [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
939 // CHECK2-NEXT:    [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
940 // CHECK2-NEXT:    [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
941 // CHECK2-NEXT:    [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
942 // CHECK2-NEXT:    store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
943 // CHECK2-NEXT:    store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
944 // CHECK2-NEXT:    ret i32 0
945 //
946 //
947 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
948 // CHECK2-SAME: (i8* noundef [[TMP0:%.*]], i8* noundef [[TMP1:%.*]]) #[[ATTR3]] {
949 // CHECK2-NEXT:  entry:
950 // CHECK2-NEXT:    [[DOTADDR:%.*]] = alloca i8*, align 8
951 // CHECK2-NEXT:    [[DOTADDR1:%.*]] = alloca i8*, align 8
952 // CHECK2-NEXT:    store i8* [[TMP0]], i8** [[DOTADDR]], align 8
953 // CHECK2-NEXT:    store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
954 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
955 // CHECK2-NEXT:    [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
956 // CHECK2-NEXT:    [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
957 // CHECK2-NEXT:    [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
958 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
959 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
960 // CHECK2-NEXT:    [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
961 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
962 // CHECK2-NEXT:    [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
963 // CHECK2-NEXT:    [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
964 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
965 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
966 // CHECK2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
967 // CHECK2-NEXT:    [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
968 // CHECK2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
969 // CHECK2-NEXT:    [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
970 // CHECK2-NEXT:    [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
971 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
972 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
973 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
974 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[TMP11]], align 4
975 // CHECK2-NEXT:    [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
976 // CHECK2-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
977 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
978 // CHECK2:       omp.arraycpy.body:
979 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
980 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
981 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
982 // CHECK2-NEXT:    [[CONV:%.*]] = sext i8 [[TMP22]] to i32
983 // CHECK2-NEXT:    [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
984 // CHECK2-NEXT:    [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
985 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
986 // CHECK2-NEXT:    [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
987 // CHECK2-NEXT:    store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
988 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
989 // CHECK2-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
990 // CHECK2-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
991 // CHECK2-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
992 // CHECK2:       omp.arraycpy.done5:
993 // CHECK2-NEXT:    ret void
994 //
995