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 sections reduction(task, +: argc, argv[0:10][0:argc])
15 {
16 #pragma omp task in_reduction(+: argc, argv[0:10][0:argc])
17 ;
18 }
19 }
20
21
22
23 // Init firstprivate copy of argc
24
25 // Init firstprivate copy of argv[0:10][0:argc]
26
27 // Register task reduction.
28
29
30
31
32
33
34
35
36
37 #endif
38 // CHECK1-LABEL: define {{[^@]+}}@main
39 // CHECK1-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
40 // CHECK1-NEXT: entry:
41 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
42 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
43 // CHECK1-NEXT: 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: [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
58 // CHECK1-NEXT: [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
59 // CHECK1-NEXT: [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
60 // CHECK1-NEXT: [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
61 // CHECK1-NEXT: [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
62 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
63 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
64 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
65 // CHECK1-NEXT: [[TMP:%.*]] = alloca i8**, align 8
66 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
67 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
68 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
69 // CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
70 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
71 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
72 // CHECK1-NEXT: [[_TMP24:%.*]] = alloca i8, align 1
73 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
74 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
75 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
76 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
77 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
78 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
79 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
80 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
81 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
82 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
83 // CHECK1-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
84 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
85 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
86 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
87 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
88 // CHECK1-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
89 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
90 // CHECK1-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
91 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
92 // CHECK1-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
93 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
94 // CHECK1-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
95 // CHECK1-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
96 // CHECK1-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
97 // CHECK1-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
98 // CHECK1-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
99 // CHECK1-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
100 // CHECK1-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
101 // CHECK1-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
102 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
103 // CHECK1-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
104 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
105 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
106 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
107 // CHECK1: omp.arrayinit.body:
108 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
109 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
110 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
111 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
112 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
113 // CHECK1: omp.arrayinit.done:
114 // CHECK1-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
115 // CHECK1-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
116 // CHECK1-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
117 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
118 // CHECK1-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
119 // CHECK1-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
120 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
121 // CHECK1-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
122 // CHECK1-NEXT: store i8* [[TMP21]], i8** [[_TMP5]], align 8
123 // 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
124 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
125 // CHECK1-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
126 // CHECK1-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
127 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
128 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
129 // CHECK1-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
130 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
131 // CHECK1-NEXT: store i64 4, i64* [[TMP26]], align 8
132 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
133 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
134 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
135 // CHECK1-NEXT: store i8* null, i8** [[TMP28]], align 8
136 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
137 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
138 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
139 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
140 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
141 // 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
142 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
143 // CHECK1-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
144 // CHECK1-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
145 // CHECK1-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
146 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
147 // CHECK1-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
148 // CHECK1-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
149 // CHECK1-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
150 // CHECK1-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
151 // CHECK1-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
152 // CHECK1-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
153 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
154 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
155 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
156 // CHECK1-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
157 // CHECK1-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
158 // CHECK1-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
159 // CHECK1-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
160 // CHECK1-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
161 // CHECK1-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
162 // CHECK1-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
163 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
164 // CHECK1-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
165 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
166 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
167 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
168 // CHECK1-NEXT: store i8* null, i8** [[TMP48]], align 8
169 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
170 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
171 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
172 // CHECK1-NEXT: store i32 1, i32* [[TMP50]], align 8
173 // CHECK1-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
174 // CHECK1-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
175 // CHECK1-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
176 // CHECK1-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
177 // CHECK1-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
178 // CHECK1-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
179 // CHECK1-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
180 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
181 // CHECK1-NEXT: [[TMP57:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
182 // CHECK1-NEXT: [[TMP58:%.*]] = icmp slt i32 [[TMP57]], 0
183 // CHECK1-NEXT: [[TMP59:%.*]] = select i1 [[TMP58]], i32 [[TMP57]], i32 0
184 // CHECK1-NEXT: store i32 [[TMP59]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
185 // CHECK1-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
186 // CHECK1-NEXT: store i32 [[TMP60]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
187 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
188 // CHECK1: omp.inner.for.cond:
189 // CHECK1-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
190 // CHECK1-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
191 // CHECK1-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP61]], [[TMP62]]
192 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
193 // CHECK1: omp.inner.for.body:
194 // CHECK1-NEXT: [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
195 // CHECK1-NEXT: switch i32 [[TMP63]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
196 // CHECK1-NEXT: i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
197 // CHECK1-NEXT: ]
198 // CHECK1: .omp.sections.case:
199 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
200 // CHECK1-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP64]], align 8
201 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
202 // CHECK1-NEXT: store i32* [[ARGC1]], i32** [[TMP65]], align 8
203 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
204 // CHECK1-NEXT: [[TMP67:%.*]] = load i8**, i8*** [[TMP]], align 8
205 // CHECK1-NEXT: store i8** [[TMP67]], i8*** [[TMP66]], align 8
206 // CHECK1-NEXT: [[TMP68:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
207 // CHECK1-NEXT: [[TMP69:%.*]] = load i32, i32* [[TMP68]], align 4
208 // CHECK1-NEXT: [[TMP70:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP69]], 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*)*))
209 // CHECK1-NEXT: [[TMP71:%.*]] = bitcast i8* [[TMP70]] to %struct.kmp_task_t_with_privates*
210 // CHECK1-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 0
211 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP72]], i32 0, i32 0
212 // CHECK1-NEXT: [[TMP74:%.*]] = load i8*, i8** [[TMP73]], align 8
213 // CHECK1-NEXT: [[TMP75:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
214 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP74]], i8* align 8 [[TMP75]], i64 24, i1 false)
215 // CHECK1-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 1
216 // CHECK1-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP76]], i32 0, i32 0
217 // CHECK1-NEXT: [[TMP78:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
218 // CHECK1-NEXT: store i8* [[TMP78]], i8** [[TMP77]], align 8
219 // CHECK1-NEXT: [[TMP79:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
220 // CHECK1-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP79]], align 4
221 // CHECK1-NEXT: [[TMP81:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP80]], i8* [[TMP70]])
222 // CHECK1-NEXT: br label [[DOTOMP_SECTIONS_EXIT]]
223 // CHECK1: .omp.sections.exit:
224 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
225 // CHECK1: omp.inner.for.inc:
226 // CHECK1-NEXT: [[TMP82:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
227 // CHECK1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP82]], 1
228 // CHECK1-NEXT: store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
229 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
230 // CHECK1: omp.inner.for.end:
231 // CHECK1-NEXT: [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
232 // CHECK1-NEXT: [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
233 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]])
234 // CHECK1-NEXT: [[TMP85:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
235 // CHECK1-NEXT: [[TMP86:%.*]] = load i32, i32* [[TMP85]], align 4
236 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP86]], i32 1)
237 // CHECK1-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
238 // CHECK1-NEXT: [[TMP88:%.*]] = bitcast i32* [[ARGC1]] to i8*
239 // CHECK1-NEXT: store i8* [[TMP88]], i8** [[TMP87]], align 8
240 // CHECK1-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
241 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP89]], align 8
242 // CHECK1-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
243 // CHECK1-NEXT: [[TMP91:%.*]] = inttoptr i64 [[TMP11]] to i8*
244 // CHECK1-NEXT: store i8* [[TMP91]], i8** [[TMP90]], align 8
245 // CHECK1-NEXT: [[TMP92:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
246 // CHECK1-NEXT: [[TMP93:%.*]] = load i32, i32* [[TMP92]], align 4
247 // CHECK1-NEXT: [[TMP94:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
248 // CHECK1-NEXT: [[TMP95:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP93]], i32 2, i64 24, i8* [[TMP94]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
249 // CHECK1-NEXT: switch i32 [[TMP95]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
250 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
251 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
252 // CHECK1-NEXT: ]
253 // CHECK1: .omp.reduction.case1:
254 // CHECK1-NEXT: [[TMP96:%.*]] = load i32, i32* [[TMP0]], align 4
255 // CHECK1-NEXT: [[TMP97:%.*]] = load i32, i32* [[ARGC1]], align 4
256 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP96]], [[TMP97]]
257 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
258 // CHECK1-NEXT: [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
259 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP98]]
260 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
261 // CHECK1: omp.arraycpy.body:
262 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
263 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
264 // CHECK1-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
265 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP99]] to i32
266 // CHECK1-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
267 // CHECK1-NEXT: [[CONV13:%.*]] = sext i8 [[TMP100]] to i32
268 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
269 // CHECK1-NEXT: [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
270 // CHECK1-NEXT: store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
271 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
272 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
273 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP98]]
274 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
275 // CHECK1: omp.arraycpy.done18:
276 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
277 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
278 // CHECK1: .omp.reduction.case2:
279 // CHECK1-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC1]], align 4
280 // CHECK1-NEXT: [[TMP102:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP101]] monotonic, align 4
281 // CHECK1-NEXT: [[TMP103:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
282 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP103]]
283 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
284 // CHECK1: omp.arraycpy.body20:
285 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
286 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
287 // CHECK1-NEXT: [[TMP104:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
288 // CHECK1-NEXT: [[CONV23:%.*]] = sext i8 [[TMP104]] to i32
289 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
290 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
291 // CHECK1: atomic_cont:
292 // CHECK1-NEXT: [[TMP105:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP110:%.*]], [[ATOMIC_CONT]] ]
293 // CHECK1-NEXT: store i8 [[TMP105]], i8* [[_TMP24]], align 1
294 // CHECK1-NEXT: [[TMP106:%.*]] = load i8, i8* [[_TMP24]], align 1
295 // CHECK1-NEXT: [[CONV25:%.*]] = sext i8 [[TMP106]] to i32
296 // CHECK1-NEXT: [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
297 // CHECK1-NEXT: [[CONV26:%.*]] = sext i8 [[TMP107]] to i32
298 // CHECK1-NEXT: [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
299 // CHECK1-NEXT: [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
300 // CHECK1-NEXT: store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
301 // CHECK1-NEXT: [[TMP108:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
302 // CHECK1-NEXT: [[TMP109:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP105]], i8 [[TMP108]] monotonic monotonic, align 1
303 // CHECK1-NEXT: [[TMP110]] = extractvalue { i8, i1 } [[TMP109]], 0
304 // CHECK1-NEXT: [[TMP111:%.*]] = extractvalue { i8, i1 } [[TMP109]], 1
305 // CHECK1-NEXT: br i1 [[TMP111]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
306 // CHECK1: atomic_exit:
307 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
308 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
309 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP103]]
310 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
311 // CHECK1: omp.arraycpy.done32:
312 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
313 // CHECK1: .omp.reduction.default:
314 // CHECK1-NEXT: [[TMP112:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
315 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP112]])
316 // CHECK1-NEXT: ret void
317 //
318 //
319 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
320 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
321 // CHECK1-NEXT: entry:
322 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
323 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
324 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
325 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
326 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
327 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
328 // CHECK1-NEXT: store i32 0, i32* [[TMP3]], align 8
329 // CHECK1-NEXT: ret void
330 //
331 //
332 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
333 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
334 // CHECK1-NEXT: entry:
335 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
336 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
337 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
338 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
339 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
340 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
341 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
342 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
343 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
344 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
345 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
346 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
347 // CHECK1-NEXT: ret void
348 //
349 //
350 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
351 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
352 // CHECK1-NEXT: entry:
353 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
354 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
355 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
356 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
357 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
358 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
359 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
360 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
361 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
362 // CHECK1: omp.arrayinit.body:
363 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
364 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
365 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
366 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
367 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
368 // CHECK1: omp.arrayinit.done:
369 // CHECK1-NEXT: ret void
370 //
371 //
372 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
373 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
374 // CHECK1-NEXT: entry:
375 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
376 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
377 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
378 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
379 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
380 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
381 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
382 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
383 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
384 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
385 // CHECK1: omp.arraycpy.body:
386 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
387 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
388 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
389 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
390 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
391 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
392 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
393 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
394 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
395 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
396 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
397 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
398 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
399 // CHECK1: omp.arraycpy.done4:
400 // CHECK1-NEXT: ret void
401 //
402 //
403 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
404 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
405 // CHECK1-NEXT: entry:
406 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
407 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
408 // CHECK1-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
409 // CHECK1-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
410 // CHECK1-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
411 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
412 // CHECK1-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
413 // CHECK1-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
414 // CHECK1-NEXT: ret void
415 //
416 //
417 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
418 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
419 // CHECK1-NEXT: entry:
420 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
421 // CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
422 // CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
423 // CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
424 // CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
425 // CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
426 // CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
427 // CHECK1-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
428 // CHECK1-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
429 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
430 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
431 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
432 // CHECK1-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
433 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
434 // CHECK1-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
435 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
436 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
437 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
438 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
439 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
440 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
441 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
442 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
443 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
444 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
445 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
446 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
447 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
448 // CHECK1-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
449 // CHECK1-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
450 // 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
451 // CHECK1-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
452 // CHECK1-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
453 // CHECK1-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
454 // CHECK1-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
455 // CHECK1-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
456 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
457 // CHECK1-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
458 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
459 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
460 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
461 // CHECK1-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
462 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
463 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
464 // CHECK1-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
465 // CHECK1-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
466 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
467 // CHECK1-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
468 // CHECK1-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
469 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
470 // CHECK1-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
471 // CHECK1-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
472 // CHECK1-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
473 // CHECK1-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
474 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
475 // CHECK1-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
476 // CHECK1-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
477 // CHECK1-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
478 // CHECK1-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
479 // CHECK1-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
480 // CHECK1-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
481 // CHECK1-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
482 // CHECK1-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
483 // CHECK1-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
484 // CHECK1-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
485 // CHECK1-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
486 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
487 // CHECK1-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
488 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
489 // CHECK1-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
490 // CHECK1-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
491 // CHECK1-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
492 // CHECK1-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
493 // CHECK1-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
494 // CHECK1-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
495 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
496 // CHECK1-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
497 // CHECK1-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
498 // CHECK1-NEXT: ret i32 0
499 //
500 //
501 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
502 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
503 // CHECK1-NEXT: entry:
504 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
505 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
506 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
507 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
508 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
509 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
510 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
511 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
512 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
513 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
514 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
515 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
516 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
517 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
518 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
519 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
520 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
521 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
522 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
523 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
524 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
525 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
526 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
527 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
528 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
529 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
530 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
531 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
532 // CHECK1: omp.arraycpy.body:
533 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
534 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
535 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
536 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
537 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
538 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
539 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
540 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
541 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
542 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
543 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
544 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
545 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
546 // CHECK1: omp.arraycpy.done5:
547 // CHECK1-NEXT: ret void
548 //
549 //
550 // CHECK2-LABEL: define {{[^@]+}}@main
551 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
552 // CHECK2-NEXT: entry:
553 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
554 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
555 // CHECK2-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
556 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
557 // CHECK2-NEXT: [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
558 // 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]])
559 // CHECK2-NEXT: ret i32 0
560 //
561 //
562 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
563 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
564 // CHECK2-NEXT: entry:
565 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
566 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
567 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
568 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
569 // CHECK2-NEXT: [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
570 // CHECK2-NEXT: [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
571 // CHECK2-NEXT: [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
572 // CHECK2-NEXT: [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
573 // CHECK2-NEXT: [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
574 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
575 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
576 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
577 // CHECK2-NEXT: [[TMP:%.*]] = alloca i8**, align 8
578 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
579 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
580 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
581 // CHECK2-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
582 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
583 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
584 // CHECK2-NEXT: [[_TMP24:%.*]] = alloca i8, align 1
585 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
586 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
587 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
588 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
589 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
590 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
591 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
592 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
593 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
594 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
595 // CHECK2-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
596 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
597 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
598 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
599 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
600 // CHECK2-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
601 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
602 // CHECK2-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
603 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
604 // CHECK2-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
605 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
606 // CHECK2-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
607 // CHECK2-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
608 // CHECK2-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
609 // CHECK2-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
610 // CHECK2-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
611 // CHECK2-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
612 // CHECK2-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
613 // CHECK2-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
614 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
615 // CHECK2-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
616 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
617 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
618 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
619 // CHECK2: omp.arrayinit.body:
620 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
621 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
622 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
623 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
624 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
625 // CHECK2: omp.arrayinit.done:
626 // CHECK2-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
627 // CHECK2-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
628 // CHECK2-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
629 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
630 // CHECK2-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
631 // CHECK2-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
632 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
633 // CHECK2-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
634 // CHECK2-NEXT: store i8* [[TMP21]], i8** [[_TMP5]], align 8
635 // 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
636 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
637 // CHECK2-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
638 // CHECK2-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
639 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
640 // CHECK2-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
641 // CHECK2-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
642 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
643 // CHECK2-NEXT: store i64 4, i64* [[TMP26]], align 8
644 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
645 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
646 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
647 // CHECK2-NEXT: store i8* null, i8** [[TMP28]], align 8
648 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
649 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
650 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
651 // CHECK2-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
652 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
653 // 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
654 // CHECK2-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
655 // CHECK2-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
656 // CHECK2-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
657 // CHECK2-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
658 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
659 // CHECK2-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
660 // CHECK2-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
661 // CHECK2-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
662 // CHECK2-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
663 // CHECK2-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
664 // CHECK2-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
665 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
666 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
667 // CHECK2-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
668 // CHECK2-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
669 // CHECK2-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
670 // CHECK2-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
671 // CHECK2-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
672 // CHECK2-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
673 // CHECK2-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
674 // CHECK2-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
675 // CHECK2-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
676 // CHECK2-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
677 // CHECK2-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
678 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
679 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
680 // CHECK2-NEXT: store i8* null, i8** [[TMP48]], align 8
681 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
682 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
683 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
684 // CHECK2-NEXT: store i32 1, i32* [[TMP50]], align 8
685 // CHECK2-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
686 // CHECK2-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
687 // CHECK2-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
688 // CHECK2-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
689 // CHECK2-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
690 // CHECK2-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
691 // CHECK2-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
692 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
693 // CHECK2-NEXT: [[TMP57:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
694 // CHECK2-NEXT: [[TMP58:%.*]] = icmp slt i32 [[TMP57]], 0
695 // CHECK2-NEXT: [[TMP59:%.*]] = select i1 [[TMP58]], i32 [[TMP57]], i32 0
696 // CHECK2-NEXT: store i32 [[TMP59]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
697 // CHECK2-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
698 // CHECK2-NEXT: store i32 [[TMP60]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
699 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
700 // CHECK2: omp.inner.for.cond:
701 // CHECK2-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
702 // CHECK2-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
703 // CHECK2-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP61]], [[TMP62]]
704 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
705 // CHECK2: omp.inner.for.body:
706 // CHECK2-NEXT: [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
707 // CHECK2-NEXT: switch i32 [[TMP63]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
708 // CHECK2-NEXT: i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
709 // CHECK2-NEXT: ]
710 // CHECK2: .omp.sections.case:
711 // CHECK2-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
712 // CHECK2-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP64]], align 8
713 // CHECK2-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
714 // CHECK2-NEXT: store i32* [[ARGC1]], i32** [[TMP65]], align 8
715 // CHECK2-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
716 // CHECK2-NEXT: [[TMP67:%.*]] = load i8**, i8*** [[TMP]], align 8
717 // CHECK2-NEXT: store i8** [[TMP67]], i8*** [[TMP66]], align 8
718 // CHECK2-NEXT: [[TMP68:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
719 // CHECK2-NEXT: [[TMP69:%.*]] = load i32, i32* [[TMP68]], align 4
720 // CHECK2-NEXT: [[TMP70:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP69]], 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*)*))
721 // CHECK2-NEXT: [[TMP71:%.*]] = bitcast i8* [[TMP70]] to %struct.kmp_task_t_with_privates*
722 // CHECK2-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 0
723 // CHECK2-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP72]], i32 0, i32 0
724 // CHECK2-NEXT: [[TMP74:%.*]] = load i8*, i8** [[TMP73]], align 8
725 // CHECK2-NEXT: [[TMP75:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
726 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP74]], i8* align 8 [[TMP75]], i64 24, i1 false)
727 // CHECK2-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP71]], i32 0, i32 1
728 // CHECK2-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP76]], i32 0, i32 0
729 // CHECK2-NEXT: [[TMP78:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
730 // CHECK2-NEXT: store i8* [[TMP78]], i8** [[TMP77]], align 8
731 // CHECK2-NEXT: [[TMP79:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
732 // CHECK2-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP79]], align 4
733 // CHECK2-NEXT: [[TMP81:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP80]], i8* [[TMP70]])
734 // CHECK2-NEXT: br label [[DOTOMP_SECTIONS_EXIT]]
735 // CHECK2: .omp.sections.exit:
736 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
737 // CHECK2: omp.inner.for.inc:
738 // CHECK2-NEXT: [[TMP82:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
739 // CHECK2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP82]], 1
740 // CHECK2-NEXT: store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
741 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
742 // CHECK2: omp.inner.for.end:
743 // CHECK2-NEXT: [[TMP83:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
744 // CHECK2-NEXT: [[TMP84:%.*]] = load i32, i32* [[TMP83]], align 4
745 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP84]])
746 // CHECK2-NEXT: [[TMP85:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
747 // CHECK2-NEXT: [[TMP86:%.*]] = load i32, i32* [[TMP85]], align 4
748 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP86]], i32 1)
749 // CHECK2-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
750 // CHECK2-NEXT: [[TMP88:%.*]] = bitcast i32* [[ARGC1]] to i8*
751 // CHECK2-NEXT: store i8* [[TMP88]], i8** [[TMP87]], align 8
752 // CHECK2-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
753 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP89]], align 8
754 // CHECK2-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
755 // CHECK2-NEXT: [[TMP91:%.*]] = inttoptr i64 [[TMP11]] to i8*
756 // CHECK2-NEXT: store i8* [[TMP91]], i8** [[TMP90]], align 8
757 // CHECK2-NEXT: [[TMP92:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
758 // CHECK2-NEXT: [[TMP93:%.*]] = load i32, i32* [[TMP92]], align 4
759 // CHECK2-NEXT: [[TMP94:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
760 // CHECK2-NEXT: [[TMP95:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP93]], i32 2, i64 24, i8* [[TMP94]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
761 // CHECK2-NEXT: switch i32 [[TMP95]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
762 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
763 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
764 // CHECK2-NEXT: ]
765 // CHECK2: .omp.reduction.case1:
766 // CHECK2-NEXT: [[TMP96:%.*]] = load i32, i32* [[TMP0]], align 4
767 // CHECK2-NEXT: [[TMP97:%.*]] = load i32, i32* [[ARGC1]], align 4
768 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP96]], [[TMP97]]
769 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
770 // CHECK2-NEXT: [[TMP98:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
771 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP98]]
772 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
773 // CHECK2: omp.arraycpy.body:
774 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
775 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
776 // CHECK2-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
777 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP99]] to i32
778 // CHECK2-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
779 // CHECK2-NEXT: [[CONV13:%.*]] = sext i8 [[TMP100]] to i32
780 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
781 // CHECK2-NEXT: [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
782 // CHECK2-NEXT: store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
783 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
784 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
785 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP98]]
786 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
787 // CHECK2: omp.arraycpy.done18:
788 // CHECK2-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP93]], [8 x i32]* @.gomp_critical_user_.reduction.var)
789 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
790 // CHECK2: .omp.reduction.case2:
791 // CHECK2-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC1]], align 4
792 // CHECK2-NEXT: [[TMP102:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP101]] monotonic, align 4
793 // CHECK2-NEXT: [[TMP103:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
794 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP103]]
795 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
796 // CHECK2: omp.arraycpy.body20:
797 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
798 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
799 // CHECK2-NEXT: [[TMP104:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
800 // CHECK2-NEXT: [[CONV23:%.*]] = sext i8 [[TMP104]] to i32
801 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
802 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
803 // CHECK2: atomic_cont:
804 // CHECK2-NEXT: [[TMP105:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP110:%.*]], [[ATOMIC_CONT]] ]
805 // CHECK2-NEXT: store i8 [[TMP105]], i8* [[_TMP24]], align 1
806 // CHECK2-NEXT: [[TMP106:%.*]] = load i8, i8* [[_TMP24]], align 1
807 // CHECK2-NEXT: [[CONV25:%.*]] = sext i8 [[TMP106]] to i32
808 // CHECK2-NEXT: [[TMP107:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
809 // CHECK2-NEXT: [[CONV26:%.*]] = sext i8 [[TMP107]] to i32
810 // CHECK2-NEXT: [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
811 // CHECK2-NEXT: [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
812 // CHECK2-NEXT: store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
813 // CHECK2-NEXT: [[TMP108:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
814 // CHECK2-NEXT: [[TMP109:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP105]], i8 [[TMP108]] monotonic monotonic, align 1
815 // CHECK2-NEXT: [[TMP110]] = extractvalue { i8, i1 } [[TMP109]], 0
816 // CHECK2-NEXT: [[TMP111:%.*]] = extractvalue { i8, i1 } [[TMP109]], 1
817 // CHECK2-NEXT: br i1 [[TMP111]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
818 // CHECK2: atomic_exit:
819 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
820 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
821 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP103]]
822 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
823 // CHECK2: omp.arraycpy.done32:
824 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
825 // CHECK2: .omp.reduction.default:
826 // CHECK2-NEXT: [[TMP112:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
827 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP112]])
828 // CHECK2-NEXT: ret void
829 //
830 //
831 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
832 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
833 // CHECK2-NEXT: entry:
834 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
835 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
836 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
837 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
838 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
839 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
840 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
841 // CHECK2-NEXT: ret void
842 //
843 //
844 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
845 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
846 // CHECK2-NEXT: entry:
847 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
848 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
849 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
850 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
851 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
852 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
853 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
854 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
855 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
856 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
857 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
858 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
859 // CHECK2-NEXT: ret void
860 //
861 //
862 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
863 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
864 // CHECK2-NEXT: entry:
865 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
866 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
867 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
868 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
869 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
870 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
871 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
872 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
873 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
874 // CHECK2: omp.arrayinit.body:
875 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
876 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
877 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
878 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
879 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
880 // CHECK2: omp.arrayinit.done:
881 // CHECK2-NEXT: ret void
882 //
883 //
884 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
885 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
886 // CHECK2-NEXT: entry:
887 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
888 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
889 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
890 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
891 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
892 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
893 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
894 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
895 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
896 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
897 // CHECK2: omp.arraycpy.body:
898 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
899 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
900 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
901 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
902 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
903 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
904 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
905 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
906 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
907 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
908 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
909 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
910 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
911 // CHECK2: omp.arraycpy.done4:
912 // CHECK2-NEXT: ret void
913 //
914 //
915 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
916 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
917 // CHECK2-NEXT: entry:
918 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
919 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
920 // CHECK2-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
921 // CHECK2-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
922 // CHECK2-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
923 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
924 // CHECK2-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
925 // CHECK2-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
926 // CHECK2-NEXT: ret void
927 //
928 //
929 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
930 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
931 // CHECK2-NEXT: entry:
932 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
933 // CHECK2-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
934 // CHECK2-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
935 // CHECK2-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
936 // CHECK2-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
937 // CHECK2-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
938 // CHECK2-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
939 // CHECK2-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
940 // CHECK2-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
941 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
942 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
943 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
944 // CHECK2-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
945 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
946 // CHECK2-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
947 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
948 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
949 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
950 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
951 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
952 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
953 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
954 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
955 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
956 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
957 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
958 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
959 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
960 // CHECK2-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
961 // CHECK2-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
962 // 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
963 // CHECK2-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
964 // CHECK2-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
965 // CHECK2-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
966 // CHECK2-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
967 // CHECK2-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
968 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
969 // CHECK2-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
970 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
971 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
972 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
973 // CHECK2-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
974 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
975 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
976 // CHECK2-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
977 // CHECK2-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
978 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
979 // CHECK2-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
980 // CHECK2-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
981 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
982 // CHECK2-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
983 // CHECK2-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
984 // CHECK2-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
985 // CHECK2-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
986 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
987 // CHECK2-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
988 // CHECK2-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
989 // CHECK2-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
990 // CHECK2-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
991 // CHECK2-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
992 // CHECK2-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
993 // CHECK2-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
994 // CHECK2-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
995 // CHECK2-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
996 // CHECK2-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
997 // CHECK2-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
998 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
999 // CHECK2-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1000 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1001 // CHECK2-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1002 // CHECK2-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1003 // CHECK2-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1004 // CHECK2-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1005 // CHECK2-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1006 // CHECK2-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1007 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1008 // CHECK2-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1009 // CHECK2-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1010 // CHECK2-NEXT: ret i32 0
1011 //
1012 //
1013 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1014 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1015 // CHECK2-NEXT: entry:
1016 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1017 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1018 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1019 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1020 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1021 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1022 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1023 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1024 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1025 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1026 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1027 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1028 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1029 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1030 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1031 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1032 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1033 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1034 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1035 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1036 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1037 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1038 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1039 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1040 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1041 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1042 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1043 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1044 // CHECK2: omp.arraycpy.body:
1045 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1046 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1047 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1048 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1049 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1050 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1051 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1052 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1053 // CHECK2-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1054 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1055 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1056 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1057 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1058 // CHECK2: omp.arraycpy.done5:
1059 // CHECK2-NEXT: ret void
1060 //
1061