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
main(int argc,char ** argv)13 int main(int argc, char **argv) {
14 #pragma omp parallel for reduction(task, +: argc, argv[0:10][0:argc])
15 for (long long i = 0; i < 10; ++i) {
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_IV:%.*]] = alloca i64, align 8
58 // CHECK1-NEXT: [[TMP:%.*]] = alloca i64, align 8
59 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
60 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
61 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
62 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
63 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
64 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
65 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
66 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
67 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
68 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
69 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
70 // CHECK1-NEXT: [[I:%.*]] = alloca i64, align 8
71 // CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
72 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
73 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
74 // CHECK1-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
75 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
76 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
77 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
78 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
79 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
80 // CHECK1-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
81 // CHECK1-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
82 // CHECK1-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
83 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
84 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
85 // CHECK1-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
86 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
87 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
88 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
89 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
90 // CHECK1-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
91 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
92 // CHECK1-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
93 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
94 // CHECK1-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
95 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
96 // CHECK1-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
97 // CHECK1-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
98 // CHECK1-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
99 // CHECK1-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
100 // CHECK1-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
101 // CHECK1-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
102 // CHECK1-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
103 // CHECK1-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
104 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
105 // CHECK1-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
106 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
107 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
108 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
109 // CHECK1: omp.arrayinit.body:
110 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
111 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
112 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
113 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
114 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
115 // CHECK1: omp.arrayinit.done:
116 // CHECK1-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
117 // CHECK1-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
118 // CHECK1-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
119 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
120 // CHECK1-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
121 // CHECK1-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
122 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
123 // CHECK1-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
124 // CHECK1-NEXT: store i8* [[TMP21]], i8** [[_TMP6]], align 8
125 // 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
126 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
127 // CHECK1-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
128 // CHECK1-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
129 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
130 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
131 // CHECK1-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
132 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
133 // CHECK1-NEXT: store i64 4, i64* [[TMP26]], align 8
134 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
135 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
136 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
137 // CHECK1-NEXT: store i8* null, i8** [[TMP28]], align 8
138 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
139 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
140 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
141 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
142 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
143 // CHECK1-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
144 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
145 // CHECK1-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
146 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
147 // CHECK1-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
148 // CHECK1-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
149 // CHECK1-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
150 // CHECK1-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
151 // CHECK1-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP36]]
152 // CHECK1-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
153 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
154 // CHECK1-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
155 // CHECK1-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN10]]
156 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
157 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
158 // CHECK1-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP39]], align 8
159 // CHECK1-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
160 // CHECK1-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
161 // CHECK1-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
162 // CHECK1-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
163 // CHECK1-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
164 // CHECK1-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
165 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
166 // CHECK1-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
167 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
168 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
169 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
170 // CHECK1-NEXT: store i8* null, i8** [[TMP48]], align 8
171 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
172 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
173 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
174 // CHECK1-NEXT: store i32 1, i32* [[TMP50]], align 8
175 // CHECK1-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
176 // CHECK1-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
177 // CHECK1-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
178 // CHECK1-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
179 // CHECK1-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
180 // CHECK1-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
181 // CHECK1-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
182 // CHECK1-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
183 // CHECK1-NEXT: [[TMP57:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
184 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP57]], 9
185 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
186 // CHECK1: cond.true:
187 // CHECK1-NEXT: br label [[COND_END:%.*]]
188 // CHECK1: cond.false:
189 // CHECK1-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
190 // CHECK1-NEXT: br label [[COND_END]]
191 // CHECK1: cond.end:
192 // CHECK1-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP58]], [[COND_FALSE]] ]
193 // CHECK1-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
194 // CHECK1-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
195 // CHECK1-NEXT: store i64 [[TMP59]], i64* [[DOTOMP_IV]], align 8
196 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
197 // CHECK1: omp.inner.for.cond:
198 // CHECK1-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
199 // CHECK1-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
200 // CHECK1-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP60]], [[TMP61]]
201 // CHECK1-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
202 // CHECK1: omp.inner.for.cond.cleanup:
203 // CHECK1-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
204 // CHECK1: omp.inner.for.body:
205 // CHECK1-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
206 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP62]], 1
207 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
208 // CHECK1-NEXT: store i64 [[ADD]], i64* [[I]], align 8
209 // CHECK1-NEXT: [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
210 // CHECK1-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP63]], align 8
211 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
212 // CHECK1-NEXT: store i32* [[ARGC1]], i32** [[TMP64]], align 8
213 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
214 // CHECK1-NEXT: [[TMP66:%.*]] = load i8**, i8*** [[_TMP5]], align 8
215 // CHECK1-NEXT: store i8** [[TMP66]], i8*** [[TMP65]], align 8
216 // CHECK1-NEXT: [[TMP67:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
217 // CHECK1-NEXT: [[TMP68:%.*]] = load i32, i32* [[TMP67]], align 4
218 // CHECK1-NEXT: [[TMP69:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP68]], 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*)*))
219 // CHECK1-NEXT: [[TMP70:%.*]] = bitcast i8* [[TMP69]] to %struct.kmp_task_t_with_privates*
220 // CHECK1-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP70]], i32 0, i32 0
221 // CHECK1-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP71]], i32 0, i32 0
222 // CHECK1-NEXT: [[TMP73:%.*]] = load i8*, i8** [[TMP72]], align 8
223 // CHECK1-NEXT: [[TMP74:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
224 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP73]], i8* align 8 [[TMP74]], i64 24, i1 false)
225 // CHECK1-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP70]], i32 0, i32 1
226 // CHECK1-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP75]], i32 0, i32 0
227 // CHECK1-NEXT: [[TMP77:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
228 // CHECK1-NEXT: store i8* [[TMP77]], i8** [[TMP76]], align 8
229 // CHECK1-NEXT: [[TMP78:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
230 // CHECK1-NEXT: [[TMP79:%.*]] = load i32, i32* [[TMP78]], align 4
231 // CHECK1-NEXT: [[TMP80:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP79]], i8* [[TMP69]])
232 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
233 // CHECK1: omp.body.continue:
234 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
235 // CHECK1: omp.inner.for.inc:
236 // CHECK1-NEXT: [[TMP81:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
237 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP81]], 1
238 // CHECK1-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
239 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
240 // CHECK1: omp.inner.for.end:
241 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
242 // CHECK1: omp.loop.exit:
243 // CHECK1-NEXT: [[TMP82:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
244 // CHECK1-NEXT: [[TMP83:%.*]] = load i32, i32* [[TMP82]], align 4
245 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP83]])
246 // CHECK1-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
247 // CHECK1-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
248 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP85]], i32 1)
249 // CHECK1-NEXT: [[TMP86:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
250 // CHECK1-NEXT: [[TMP87:%.*]] = bitcast i32* [[ARGC1]] to i8*
251 // CHECK1-NEXT: store i8* [[TMP87]], i8** [[TMP86]], align 8
252 // CHECK1-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
253 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP88]], align 8
254 // CHECK1-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
255 // CHECK1-NEXT: [[TMP90:%.*]] = inttoptr i64 [[TMP11]] to i8*
256 // CHECK1-NEXT: store i8* [[TMP90]], i8** [[TMP89]], align 8
257 // CHECK1-NEXT: [[TMP91:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
258 // CHECK1-NEXT: [[TMP92:%.*]] = load i32, i32* [[TMP91]], align 4
259 // CHECK1-NEXT: [[TMP93:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
260 // CHECK1-NEXT: [[TMP94:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP92]], i32 2, i64 24, i8* [[TMP93]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
261 // CHECK1-NEXT: switch i32 [[TMP94]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
262 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
263 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
264 // CHECK1-NEXT: ]
265 // CHECK1: .omp.reduction.case1:
266 // CHECK1-NEXT: [[TMP95:%.*]] = load i32, i32* [[TMP0]], align 4
267 // CHECK1-NEXT: [[TMP96:%.*]] = load i32, i32* [[ARGC1]], align 4
268 // CHECK1-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP95]], [[TMP96]]
269 // CHECK1-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
270 // CHECK1-NEXT: [[TMP97:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
271 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP97]]
272 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
273 // CHECK1: omp.arraycpy.body:
274 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
275 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
276 // CHECK1-NEXT: [[TMP98:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
277 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP98]] to i32
278 // CHECK1-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
279 // CHECK1-NEXT: [[CONV17:%.*]] = sext i8 [[TMP99]] to i32
280 // CHECK1-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
281 // CHECK1-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
282 // CHECK1-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
283 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
284 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
285 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP97]]
286 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
287 // CHECK1: omp.arraycpy.done22:
288 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP92]], [8 x i32]* @.gomp_critical_user_.reduction.var)
289 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
290 // CHECK1: .omp.reduction.case2:
291 // CHECK1-NEXT: [[TMP100:%.*]] = load i32, i32* [[ARGC1]], align 4
292 // CHECK1-NEXT: [[TMP101:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP100]] monotonic, align 4
293 // CHECK1-NEXT: [[TMP102:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
294 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP102]]
295 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
296 // CHECK1: omp.arraycpy.body24:
297 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
298 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
299 // CHECK1-NEXT: [[TMP103:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
300 // CHECK1-NEXT: [[CONV27:%.*]] = sext i8 [[TMP103]] to i32
301 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
302 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
303 // CHECK1: atomic_cont:
304 // CHECK1-NEXT: [[TMP104:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP109:%.*]], [[ATOMIC_CONT]] ]
305 // CHECK1-NEXT: store i8 [[TMP104]], i8* [[_TMP28]], align 1
306 // CHECK1-NEXT: [[TMP105:%.*]] = load i8, i8* [[_TMP28]], align 1
307 // CHECK1-NEXT: [[CONV29:%.*]] = sext i8 [[TMP105]] to i32
308 // CHECK1-NEXT: [[TMP106:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
309 // CHECK1-NEXT: [[CONV30:%.*]] = sext i8 [[TMP106]] to i32
310 // CHECK1-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
311 // CHECK1-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
312 // CHECK1-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
313 // CHECK1-NEXT: [[TMP107:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
314 // CHECK1-NEXT: [[TMP108:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP104]], i8 [[TMP107]] monotonic monotonic, align 1
315 // CHECK1-NEXT: [[TMP109]] = extractvalue { i8, i1 } [[TMP108]], 0
316 // CHECK1-NEXT: [[TMP110:%.*]] = extractvalue { i8, i1 } [[TMP108]], 1
317 // CHECK1-NEXT: br i1 [[TMP110]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
318 // CHECK1: atomic_exit:
319 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
320 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
321 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP102]]
322 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
323 // CHECK1: omp.arraycpy.done36:
324 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
325 // CHECK1: .omp.reduction.default:
326 // CHECK1-NEXT: [[TMP111:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
327 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP111]])
328 // CHECK1-NEXT: ret void
329 //
330 //
331 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
332 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
333 // CHECK1-NEXT: entry:
334 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
335 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
336 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
337 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
338 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
339 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
340 // CHECK1-NEXT: store i32 0, i32* [[TMP3]], align 8
341 // CHECK1-NEXT: ret void
342 //
343 //
344 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
345 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
346 // CHECK1-NEXT: entry:
347 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
348 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
349 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
350 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
351 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
352 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
353 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
354 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
355 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
356 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
357 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
358 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
359 // CHECK1-NEXT: ret void
360 //
361 //
362 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
363 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
364 // CHECK1-NEXT: entry:
365 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
366 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
367 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
368 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
369 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
370 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
371 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
372 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
373 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
374 // CHECK1: omp.arrayinit.body:
375 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
376 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
377 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
378 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
379 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
380 // CHECK1: omp.arrayinit.done:
381 // CHECK1-NEXT: ret void
382 //
383 //
384 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
385 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
386 // CHECK1-NEXT: entry:
387 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
388 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
389 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
390 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
391 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
392 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
393 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
394 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
395 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
396 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
397 // CHECK1: omp.arraycpy.body:
398 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
399 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
400 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
401 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
402 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
403 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
404 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
405 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
406 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
407 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
408 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
409 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
410 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
411 // CHECK1: omp.arraycpy.done4:
412 // CHECK1-NEXT: ret void
413 //
414 //
415 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
416 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
417 // CHECK1-NEXT: entry:
418 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
419 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
420 // CHECK1-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
421 // CHECK1-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
422 // CHECK1-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
423 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
424 // CHECK1-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
425 // CHECK1-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
426 // CHECK1-NEXT: ret void
427 //
428 //
429 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
430 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
431 // CHECK1-NEXT: entry:
432 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
433 // CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
434 // CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
435 // CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
436 // CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
437 // CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
438 // CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
439 // CHECK1-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
440 // CHECK1-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
441 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
442 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
443 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
444 // CHECK1-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
445 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
446 // CHECK1-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
447 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
448 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
449 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
450 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
451 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
452 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
453 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
454 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
455 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
456 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
457 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
458 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
459 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
460 // CHECK1-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
461 // CHECK1-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
462 // 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
463 // CHECK1-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
464 // CHECK1-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
465 // CHECK1-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
466 // CHECK1-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
467 // CHECK1-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
468 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
469 // CHECK1-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
470 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
471 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
472 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
473 // CHECK1-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
474 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
475 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
476 // CHECK1-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
477 // CHECK1-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
478 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
479 // CHECK1-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
480 // CHECK1-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
481 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
482 // CHECK1-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
483 // CHECK1-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
484 // CHECK1-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
485 // CHECK1-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
486 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
487 // CHECK1-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
488 // CHECK1-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
489 // CHECK1-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
490 // CHECK1-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
491 // CHECK1-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
492 // CHECK1-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
493 // CHECK1-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
494 // CHECK1-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
495 // CHECK1-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
496 // CHECK1-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
497 // CHECK1-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
498 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
499 // CHECK1-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
500 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
501 // CHECK1-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
502 // CHECK1-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
503 // CHECK1-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
504 // CHECK1-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
505 // CHECK1-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
506 // CHECK1-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
507 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
508 // CHECK1-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
509 // CHECK1-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
510 // CHECK1-NEXT: ret i32 0
511 //
512 //
513 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
514 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
515 // CHECK1-NEXT: entry:
516 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
517 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
518 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
519 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
520 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
521 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
522 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
523 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
524 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
525 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
526 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
527 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
528 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
529 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
530 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
531 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
532 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
533 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
534 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
535 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
536 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
537 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
538 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
539 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
540 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
541 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
542 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
543 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
544 // CHECK1: omp.arraycpy.body:
545 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
546 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
547 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
548 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
549 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
550 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
551 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
552 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
553 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
554 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
555 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
556 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
557 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
558 // CHECK1: omp.arraycpy.done5:
559 // CHECK1-NEXT: ret void
560 //
561 //
562 // CHECK2-LABEL: define {{[^@]+}}@main
563 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
564 // CHECK2-NEXT: entry:
565 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
566 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
567 // CHECK2-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
568 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
569 // CHECK2-NEXT: [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
570 // 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]])
571 // CHECK2-NEXT: ret i32 0
572 //
573 //
574 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
575 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
576 // CHECK2-NEXT: entry:
577 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
578 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
579 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
580 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
581 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
582 // CHECK2-NEXT: [[TMP:%.*]] = alloca i64, align 8
583 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
584 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
585 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
586 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
587 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
588 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
589 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
590 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
591 // CHECK2-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
592 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
593 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
594 // CHECK2-NEXT: [[I:%.*]] = alloca i64, align 8
595 // CHECK2-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
596 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
597 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
598 // CHECK2-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
599 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
600 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
601 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
602 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
603 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
604 // CHECK2-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
605 // CHECK2-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
606 // CHECK2-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
607 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
608 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
609 // CHECK2-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
610 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
611 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
612 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
613 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
614 // CHECK2-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
615 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
616 // CHECK2-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
617 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
618 // CHECK2-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
619 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
620 // CHECK2-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
621 // CHECK2-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
622 // CHECK2-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
623 // CHECK2-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
624 // CHECK2-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
625 // CHECK2-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
626 // CHECK2-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
627 // CHECK2-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
628 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
629 // CHECK2-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
630 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
631 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
632 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
633 // CHECK2: omp.arrayinit.body:
634 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
635 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
636 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
637 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
638 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
639 // CHECK2: omp.arrayinit.done:
640 // CHECK2-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
641 // CHECK2-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
642 // CHECK2-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
643 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
644 // CHECK2-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
645 // CHECK2-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
646 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
647 // CHECK2-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
648 // CHECK2-NEXT: store i8* [[TMP21]], i8** [[_TMP6]], align 8
649 // 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
650 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
651 // CHECK2-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
652 // CHECK2-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
653 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
654 // CHECK2-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
655 // CHECK2-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
656 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
657 // CHECK2-NEXT: store i64 4, i64* [[TMP26]], align 8
658 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
659 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
660 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
661 // CHECK2-NEXT: store i8* null, i8** [[TMP28]], align 8
662 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
663 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
664 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
665 // CHECK2-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
666 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
667 // CHECK2-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t], [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]], i64 0, i64 1
668 // CHECK2-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
669 // CHECK2-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
670 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
671 // CHECK2-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
672 // CHECK2-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
673 // CHECK2-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
674 // CHECK2-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
675 // CHECK2-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP36]]
676 // CHECK2-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
677 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
678 // CHECK2-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
679 // CHECK2-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN10]]
680 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
681 // CHECK2-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
682 // CHECK2-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP39]], align 8
683 // CHECK2-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
684 // CHECK2-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
685 // CHECK2-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
686 // CHECK2-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
687 // CHECK2-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
688 // CHECK2-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
689 // CHECK2-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
690 // CHECK2-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
691 // CHECK2-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
692 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
693 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
694 // CHECK2-NEXT: store i8* null, i8** [[TMP48]], align 8
695 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
696 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
697 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
698 // CHECK2-NEXT: store i32 1, i32* [[TMP50]], align 8
699 // CHECK2-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
700 // CHECK2-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
701 // CHECK2-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
702 // CHECK2-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
703 // CHECK2-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
704 // CHECK2-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
705 // CHECK2-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
706 // CHECK2-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
707 // CHECK2-NEXT: [[TMP57:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
708 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP57]], 9
709 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
710 // CHECK2: cond.true:
711 // CHECK2-NEXT: br label [[COND_END:%.*]]
712 // CHECK2: cond.false:
713 // CHECK2-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
714 // CHECK2-NEXT: br label [[COND_END]]
715 // CHECK2: cond.end:
716 // CHECK2-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP58]], [[COND_FALSE]] ]
717 // CHECK2-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
718 // CHECK2-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
719 // CHECK2-NEXT: store i64 [[TMP59]], i64* [[DOTOMP_IV]], align 8
720 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
721 // CHECK2: omp.inner.for.cond:
722 // CHECK2-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
723 // CHECK2-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
724 // CHECK2-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP60]], [[TMP61]]
725 // CHECK2-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
726 // CHECK2: omp.inner.for.cond.cleanup:
727 // CHECK2-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
728 // CHECK2: omp.inner.for.body:
729 // CHECK2-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
730 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP62]], 1
731 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
732 // CHECK2-NEXT: store i64 [[ADD]], i64* [[I]], align 8
733 // CHECK2-NEXT: [[TMP63:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
734 // CHECK2-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP63]], align 8
735 // CHECK2-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
736 // CHECK2-NEXT: store i32* [[ARGC1]], i32** [[TMP64]], align 8
737 // CHECK2-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
738 // CHECK2-NEXT: [[TMP66:%.*]] = load i8**, i8*** [[_TMP5]], align 8
739 // CHECK2-NEXT: store i8** [[TMP66]], i8*** [[TMP65]], align 8
740 // CHECK2-NEXT: [[TMP67:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
741 // CHECK2-NEXT: [[TMP68:%.*]] = load i32, i32* [[TMP67]], align 4
742 // CHECK2-NEXT: [[TMP69:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP68]], 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*)*))
743 // CHECK2-NEXT: [[TMP70:%.*]] = bitcast i8* [[TMP69]] to %struct.kmp_task_t_with_privates*
744 // CHECK2-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP70]], i32 0, i32 0
745 // CHECK2-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP71]], i32 0, i32 0
746 // CHECK2-NEXT: [[TMP73:%.*]] = load i8*, i8** [[TMP72]], align 8
747 // CHECK2-NEXT: [[TMP74:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
748 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP73]], i8* align 8 [[TMP74]], i64 24, i1 false)
749 // CHECK2-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP70]], i32 0, i32 1
750 // CHECK2-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP75]], i32 0, i32 0
751 // CHECK2-NEXT: [[TMP77:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
752 // CHECK2-NEXT: store i8* [[TMP77]], i8** [[TMP76]], align 8
753 // CHECK2-NEXT: [[TMP78:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
754 // CHECK2-NEXT: [[TMP79:%.*]] = load i32, i32* [[TMP78]], align 4
755 // CHECK2-NEXT: [[TMP80:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP79]], i8* [[TMP69]])
756 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
757 // CHECK2: omp.body.continue:
758 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
759 // CHECK2: omp.inner.for.inc:
760 // CHECK2-NEXT: [[TMP81:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
761 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP81]], 1
762 // CHECK2-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
763 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
764 // CHECK2: omp.inner.for.end:
765 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
766 // CHECK2: omp.loop.exit:
767 // CHECK2-NEXT: [[TMP82:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
768 // CHECK2-NEXT: [[TMP83:%.*]] = load i32, i32* [[TMP82]], align 4
769 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP83]])
770 // CHECK2-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
771 // CHECK2-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
772 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP85]], i32 1)
773 // CHECK2-NEXT: [[TMP86:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
774 // CHECK2-NEXT: [[TMP87:%.*]] = bitcast i32* [[ARGC1]] to i8*
775 // CHECK2-NEXT: store i8* [[TMP87]], i8** [[TMP86]], align 8
776 // CHECK2-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
777 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP88]], align 8
778 // CHECK2-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
779 // CHECK2-NEXT: [[TMP90:%.*]] = inttoptr i64 [[TMP11]] to i8*
780 // CHECK2-NEXT: store i8* [[TMP90]], i8** [[TMP89]], align 8
781 // CHECK2-NEXT: [[TMP91:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
782 // CHECK2-NEXT: [[TMP92:%.*]] = load i32, i32* [[TMP91]], align 4
783 // CHECK2-NEXT: [[TMP93:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
784 // CHECK2-NEXT: [[TMP94:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP92]], i32 2, i64 24, i8* [[TMP93]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
785 // CHECK2-NEXT: switch i32 [[TMP94]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
786 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
787 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
788 // CHECK2-NEXT: ]
789 // CHECK2: .omp.reduction.case1:
790 // CHECK2-NEXT: [[TMP95:%.*]] = load i32, i32* [[TMP0]], align 4
791 // CHECK2-NEXT: [[TMP96:%.*]] = load i32, i32* [[ARGC1]], align 4
792 // CHECK2-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP95]], [[TMP96]]
793 // CHECK2-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
794 // CHECK2-NEXT: [[TMP97:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
795 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP97]]
796 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
797 // CHECK2: omp.arraycpy.body:
798 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
799 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
800 // CHECK2-NEXT: [[TMP98:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
801 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP98]] to i32
802 // CHECK2-NEXT: [[TMP99:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
803 // CHECK2-NEXT: [[CONV17:%.*]] = sext i8 [[TMP99]] to i32
804 // CHECK2-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
805 // CHECK2-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
806 // CHECK2-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
807 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
808 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
809 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP97]]
810 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
811 // CHECK2: omp.arraycpy.done22:
812 // CHECK2-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB3]], i32 [[TMP92]], [8 x i32]* @.gomp_critical_user_.reduction.var)
813 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
814 // CHECK2: .omp.reduction.case2:
815 // CHECK2-NEXT: [[TMP100:%.*]] = load i32, i32* [[ARGC1]], align 4
816 // CHECK2-NEXT: [[TMP101:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP100]] monotonic, align 4
817 // CHECK2-NEXT: [[TMP102:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
818 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP102]]
819 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
820 // CHECK2: omp.arraycpy.body24:
821 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
822 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
823 // CHECK2-NEXT: [[TMP103:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
824 // CHECK2-NEXT: [[CONV27:%.*]] = sext i8 [[TMP103]] to i32
825 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
826 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
827 // CHECK2: atomic_cont:
828 // CHECK2-NEXT: [[TMP104:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP109:%.*]], [[ATOMIC_CONT]] ]
829 // CHECK2-NEXT: store i8 [[TMP104]], i8* [[_TMP28]], align 1
830 // CHECK2-NEXT: [[TMP105:%.*]] = load i8, i8* [[_TMP28]], align 1
831 // CHECK2-NEXT: [[CONV29:%.*]] = sext i8 [[TMP105]] to i32
832 // CHECK2-NEXT: [[TMP106:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
833 // CHECK2-NEXT: [[CONV30:%.*]] = sext i8 [[TMP106]] to i32
834 // CHECK2-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
835 // CHECK2-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
836 // CHECK2-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
837 // CHECK2-NEXT: [[TMP107:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
838 // CHECK2-NEXT: [[TMP108:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP104]], i8 [[TMP107]] monotonic monotonic, align 1
839 // CHECK2-NEXT: [[TMP109]] = extractvalue { i8, i1 } [[TMP108]], 0
840 // CHECK2-NEXT: [[TMP110:%.*]] = extractvalue { i8, i1 } [[TMP108]], 1
841 // CHECK2-NEXT: br i1 [[TMP110]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
842 // CHECK2: atomic_exit:
843 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
844 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
845 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP102]]
846 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
847 // CHECK2: omp.arraycpy.done36:
848 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
849 // CHECK2: .omp.reduction.default:
850 // CHECK2-NEXT: [[TMP111:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
851 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP111]])
852 // CHECK2-NEXT: ret void
853 //
854 //
855 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
856 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
857 // CHECK2-NEXT: entry:
858 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
859 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
860 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
861 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
862 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
863 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
864 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
865 // CHECK2-NEXT: ret void
866 //
867 //
868 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
869 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
870 // CHECK2-NEXT: entry:
871 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
872 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
873 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
874 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
875 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
876 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
877 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
878 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
879 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
880 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
881 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
882 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
883 // CHECK2-NEXT: ret void
884 //
885 //
886 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
887 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
888 // CHECK2-NEXT: entry:
889 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
890 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
891 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
892 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
893 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
894 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
895 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
896 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
897 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
898 // CHECK2: omp.arrayinit.body:
899 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
900 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
901 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
902 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
903 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
904 // CHECK2: omp.arrayinit.done:
905 // CHECK2-NEXT: ret void
906 //
907 //
908 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
909 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
910 // CHECK2-NEXT: entry:
911 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
912 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
913 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
914 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
915 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
916 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
917 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
918 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
919 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
920 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
921 // CHECK2: omp.arraycpy.body:
922 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
923 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
924 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
925 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
926 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
927 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
928 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
929 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
930 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
931 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
932 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
933 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
934 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
935 // CHECK2: omp.arraycpy.done4:
936 // CHECK2-NEXT: ret void
937 //
938 //
939 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
940 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
941 // CHECK2-NEXT: entry:
942 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
943 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
944 // CHECK2-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
945 // CHECK2-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
946 // CHECK2-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
947 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
948 // CHECK2-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
949 // CHECK2-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
950 // CHECK2-NEXT: ret void
951 //
952 //
953 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
954 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
955 // CHECK2-NEXT: entry:
956 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
957 // CHECK2-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
958 // CHECK2-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
959 // CHECK2-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
960 // CHECK2-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
961 // CHECK2-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
962 // CHECK2-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
963 // CHECK2-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
964 // CHECK2-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
965 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
966 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
967 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
968 // CHECK2-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
969 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
970 // CHECK2-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
971 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
972 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
973 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
974 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
975 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
976 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
977 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
978 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
979 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
980 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
981 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
982 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
983 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
984 // CHECK2-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
985 // CHECK2-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
986 // 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
987 // CHECK2-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
988 // CHECK2-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
989 // CHECK2-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
990 // CHECK2-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
991 // CHECK2-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
992 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
993 // CHECK2-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
994 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
995 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
996 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
997 // CHECK2-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
998 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
999 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
1000 // CHECK2-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
1001 // CHECK2-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
1002 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1003 // CHECK2-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
1004 // CHECK2-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
1005 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
1006 // CHECK2-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
1007 // CHECK2-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
1008 // CHECK2-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
1009 // CHECK2-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
1010 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1011 // CHECK2-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
1012 // CHECK2-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
1013 // CHECK2-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
1014 // CHECK2-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
1015 // CHECK2-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
1016 // CHECK2-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
1017 // CHECK2-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
1018 // CHECK2-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1019 // CHECK2-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
1020 // CHECK2-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1021 // CHECK2-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
1022 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
1023 // CHECK2-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1024 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1025 // CHECK2-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1026 // CHECK2-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1027 // CHECK2-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1028 // CHECK2-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1029 // CHECK2-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1030 // CHECK2-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1031 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1032 // CHECK2-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1033 // CHECK2-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1034 // CHECK2-NEXT: ret i32 0
1035 //
1036 //
1037 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1038 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1039 // CHECK2-NEXT: entry:
1040 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1041 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1042 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1043 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1044 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1045 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1046 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1047 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1048 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1049 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1050 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1051 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1052 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1053 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1054 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1055 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1056 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1057 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1058 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1059 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1060 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1061 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1062 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1063 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1064 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1065 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1066 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1067 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1068 // CHECK2: omp.arraycpy.body:
1069 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1070 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1071 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1072 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1073 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1074 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1075 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1076 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1077 // CHECK2-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1078 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1079 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1080 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1081 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1082 // CHECK2: omp.arraycpy.done5:
1083 // CHECK2-NEXT: ret void
1084 //
1085