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 target
15 #pragma omp teams distribute parallel for reduction(task, +: argc, argv[0:10][0:argc])
16 for (long long i = 0; i < 10; ++i) {
17 #pragma omp task in_reduction(+: argc, argv[0:10][0:argc])
18 ;
19 }
20 }
21
22
23
24 // Init firstprivate copy of argc
25
26 // Init firstprivate copy of argv[0:10][0:argc]
27
28 // Register task reduction.
29
30
31
32
33
34
35
36
37
38 #endif
39 // CHECK1-LABEL: define {{[^@]+}}@main
40 // CHECK1-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
41 // CHECK1-NEXT: entry:
42 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
43 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
44 // CHECK1-NEXT: [[ARGC_CASTED:%.*]] = alloca i64, align 8
45 // CHECK1-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
46 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
47 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
48 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
49 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
50 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
51 // CHECK1-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
52 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14(i64 [[TMP1]], i8** [[TMP2]]) #[[ATTR5:[0-9]+]]
53 // CHECK1-NEXT: ret i32 0
54 //
55 //
56 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14
57 // CHECK1-SAME: (i64 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
58 // CHECK1-NEXT: entry:
59 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i64, align 8
60 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
61 // CHECK1-NEXT: store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
62 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
63 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
64 // CHECK1-NEXT: [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
65 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i8** [[TMP0]])
66 // CHECK1-NEXT: ret void
67 //
68 //
69 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
70 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1]] {
71 // CHECK1-NEXT: entry:
72 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
73 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
74 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
75 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
76 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
77 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
78 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
79 // CHECK1-NEXT: [[TMP:%.*]] = alloca i8**, align 8
80 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
81 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
82 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
83 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
84 // CHECK1-NEXT: [[_TMP12:%.*]] = alloca i64, align 8
85 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
86 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
87 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
88 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
89 // CHECK1-NEXT: [[I:%.*]] = alloca i64, align 8
90 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
91 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
92 // CHECK1-NEXT: [[_TMP27:%.*]] = alloca i8, align 1
93 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
94 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
95 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
96 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
97 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
98 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
99 // CHECK1-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
100 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
101 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
102 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
103 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
104 // CHECK1-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
105 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
106 // CHECK1-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
107 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
108 // CHECK1-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
109 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
110 // CHECK1-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
111 // CHECK1-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
112 // CHECK1-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
113 // CHECK1-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
114 // CHECK1-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
115 // CHECK1-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
116 // CHECK1-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
117 // CHECK1-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
118 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
119 // CHECK1-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
120 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
121 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
122 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
123 // CHECK1: omp.arrayinit.body:
124 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
125 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
126 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
127 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
128 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
129 // CHECK1: omp.arrayinit.done:
130 // CHECK1-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
131 // CHECK1-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
132 // CHECK1-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
133 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
134 // CHECK1-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
135 // CHECK1-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
136 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
137 // CHECK1-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
138 // CHECK1-NEXT: store i8* [[TMP21]], i8** [[_TMP5]], align 8
139 // 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
140 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
141 // CHECK1-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
142 // CHECK1-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
143 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
144 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
145 // CHECK1-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
146 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
147 // CHECK1-NEXT: store i64 4, i64* [[TMP26]], align 8
148 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
149 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
150 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
151 // CHECK1-NEXT: store i8* null, i8** [[TMP28]], align 8
152 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
153 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
154 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
155 // CHECK1-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
156 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
157 // 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
158 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
159 // CHECK1-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
160 // CHECK1-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
161 // CHECK1-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
162 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
163 // CHECK1-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
164 // CHECK1-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
165 // CHECK1-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
166 // CHECK1-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
167 // CHECK1-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
168 // CHECK1-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
169 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
170 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
171 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
172 // CHECK1-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
173 // CHECK1-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
174 // CHECK1-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
175 // CHECK1-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
176 // CHECK1-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
177 // CHECK1-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
178 // CHECK1-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
179 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
180 // CHECK1-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
181 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
182 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
183 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
184 // CHECK1-NEXT: store i8* null, i8** [[TMP48]], align 8
185 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
186 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
187 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
188 // CHECK1-NEXT: store i32 1, i32* [[TMP50]], align 8
189 // CHECK1-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
190 // CHECK1-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
191 // CHECK1-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
192 // CHECK1-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
193 // CHECK1-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
194 // CHECK1-NEXT: store i64 0, i64* [[DOTOMP_COMB_LB]], align 8
195 // CHECK1-NEXT: store i64 9, i64* [[DOTOMP_COMB_UB]], align 8
196 // CHECK1-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
197 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
198 // CHECK1-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
199 // CHECK1-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
200 // CHECK1-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_COMB_LB]], i64* [[DOTOMP_COMB_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
201 // CHECK1-NEXT: [[TMP57:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
202 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP57]], 9
203 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
204 // CHECK1: cond.true:
205 // CHECK1-NEXT: br label [[COND_END:%.*]]
206 // CHECK1: cond.false:
207 // CHECK1-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
208 // CHECK1-NEXT: br label [[COND_END]]
209 // CHECK1: cond.end:
210 // CHECK1-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP58]], [[COND_FALSE]] ]
211 // CHECK1-NEXT: store i64 [[COND]], i64* [[DOTOMP_COMB_UB]], align 8
212 // CHECK1-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
213 // CHECK1-NEXT: store i64 [[TMP59]], i64* [[DOTOMP_IV]], align 8
214 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
215 // CHECK1: omp.inner.for.cond:
216 // CHECK1-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
217 // CHECK1-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
218 // CHECK1-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP60]], [[TMP61]]
219 // CHECK1-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
220 // CHECK1: omp.inner.for.body:
221 // CHECK1-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
222 // CHECK1-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
223 // CHECK1-NEXT: [[TMP64:%.*]] = load i8**, i8*** [[TMP]], align 8
224 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i8**)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP62]], i64 [[TMP63]], i32* [[ARGC1]], i8** [[TMP64]])
225 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
226 // CHECK1: omp.inner.for.inc:
227 // CHECK1-NEXT: [[TMP65:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
228 // CHECK1-NEXT: [[TMP66:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
229 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i64 [[TMP65]], [[TMP66]]
230 // CHECK1-NEXT: store i64 [[ADD]], i64* [[DOTOMP_IV]], align 8
231 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
232 // CHECK1: omp.inner.for.end:
233 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
234 // CHECK1: omp.loop.exit:
235 // CHECK1-NEXT: [[TMP67:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
236 // CHECK1-NEXT: [[TMP68:%.*]] = load i32, i32* [[TMP67]], align 4
237 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP68]])
238 // CHECK1-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
239 // CHECK1-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
240 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP70]], i32 1)
241 // CHECK1-NEXT: [[TMP71:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
242 // CHECK1-NEXT: [[TMP72:%.*]] = bitcast i32* [[ARGC1]] to i8*
243 // CHECK1-NEXT: store i8* [[TMP72]], i8** [[TMP71]], align 8
244 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
245 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP73]], align 8
246 // CHECK1-NEXT: [[TMP74:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
247 // CHECK1-NEXT: [[TMP75:%.*]] = inttoptr i64 [[TMP11]] to i8*
248 // CHECK1-NEXT: store i8* [[TMP75]], i8** [[TMP74]], align 8
249 // CHECK1-NEXT: [[TMP76:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
250 // CHECK1-NEXT: [[TMP77:%.*]] = load i32, i32* [[TMP76]], align 4
251 // CHECK1-NEXT: [[TMP78:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
252 // CHECK1-NEXT: [[TMP79:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP77]], i32 2, i64 24, i8* [[TMP78]], void (i8*, i8*)* @.omp.reduction.reduction_func.9, [8 x i32]* @.gomp_critical_user_.reduction.var)
253 // CHECK1-NEXT: switch i32 [[TMP79]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
254 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
255 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
256 // CHECK1-NEXT: ]
257 // CHECK1: .omp.reduction.case1:
258 // CHECK1-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP0]], align 4
259 // CHECK1-NEXT: [[TMP81:%.*]] = load i32, i32* [[ARGC1]], align 4
260 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP80]], [[TMP81]]
261 // CHECK1-NEXT: store i32 [[ADD14]], i32* [[TMP0]], align 4
262 // CHECK1-NEXT: [[TMP82:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
263 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP82]]
264 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
265 // CHECK1: omp.arraycpy.body:
266 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
267 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY]] ]
268 // CHECK1-NEXT: [[TMP83:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 1
269 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP83]] to i32
270 // CHECK1-NEXT: [[TMP84:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
271 // CHECK1-NEXT: [[CONV16:%.*]] = sext i8 [[TMP84]] to i32
272 // CHECK1-NEXT: [[ADD17:%.*]] = add nsw i32 [[CONV]], [[CONV16]]
273 // CHECK1-NEXT: [[CONV18:%.*]] = trunc i32 [[ADD17]] to i8
274 // CHECK1-NEXT: store i8 [[CONV18]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 1
275 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
276 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
277 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP82]]
278 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY]]
279 // CHECK1: omp.arraycpy.done21:
280 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP77]], [8 x i32]* @.gomp_critical_user_.reduction.var)
281 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
282 // CHECK1: .omp.reduction.case2:
283 // CHECK1-NEXT: [[TMP85:%.*]] = load i32, i32* [[ARGC1]], align 4
284 // CHECK1-NEXT: [[TMP86:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP85]] monotonic, align 4
285 // CHECK1-NEXT: [[TMP87:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
286 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY22:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP87]]
287 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY22]], label [[OMP_ARRAYCPY_DONE35:%.*]], label [[OMP_ARRAYCPY_BODY23:%.*]]
288 // CHECK1: omp.arraycpy.body23:
289 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST24:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT33:%.*]], [[ATOMIC_EXIT:%.*]] ]
290 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST25:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT32:%.*]], [[ATOMIC_EXIT]] ]
291 // CHECK1-NEXT: [[TMP88:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], align 1
292 // CHECK1-NEXT: [[CONV26:%.*]] = sext i8 [[TMP88]] to i32
293 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]] monotonic, align 1
294 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
295 // CHECK1: atomic_cont:
296 // CHECK1-NEXT: [[TMP89:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY23]] ], [ [[TMP94:%.*]], [[ATOMIC_CONT]] ]
297 // CHECK1-NEXT: store i8 [[TMP89]], i8* [[_TMP27]], align 1
298 // CHECK1-NEXT: [[TMP90:%.*]] = load i8, i8* [[_TMP27]], align 1
299 // CHECK1-NEXT: [[CONV28:%.*]] = sext i8 [[TMP90]] to i32
300 // CHECK1-NEXT: [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], align 1
301 // CHECK1-NEXT: [[CONV29:%.*]] = sext i8 [[TMP91]] to i32
302 // CHECK1-NEXT: [[ADD30:%.*]] = add nsw i32 [[CONV28]], [[CONV29]]
303 // CHECK1-NEXT: [[CONV31:%.*]] = trunc i32 [[ADD30]] to i8
304 // CHECK1-NEXT: store i8 [[CONV31]], i8* [[ATOMIC_TEMP]], align 1
305 // CHECK1-NEXT: [[TMP92:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
306 // CHECK1-NEXT: [[TMP93:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]], i8 [[TMP89]], i8 [[TMP92]] monotonic monotonic, align 1
307 // CHECK1-NEXT: [[TMP94]] = extractvalue { i8, i1 } [[TMP93]], 0
308 // CHECK1-NEXT: [[TMP95:%.*]] = extractvalue { i8, i1 } [[TMP93]], 1
309 // CHECK1-NEXT: br i1 [[TMP95]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
310 // CHECK1: atomic_exit:
311 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT32]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]], i32 1
312 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], i32 1
313 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE34:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT32]], [[TMP87]]
314 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE34]], label [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_BODY23]]
315 // CHECK1: omp.arraycpy.done35:
316 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
317 // CHECK1: .omp.reduction.default:
318 // CHECK1-NEXT: [[TMP96:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
319 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP96]])
320 // CHECK1-NEXT: ret void
321 //
322 //
323 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
324 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
325 // CHECK1-NEXT: entry:
326 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
327 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
328 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
329 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
330 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
331 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
332 // CHECK1-NEXT: store i32 0, i32* [[TMP3]], align 8
333 // CHECK1-NEXT: ret void
334 //
335 //
336 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
337 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
338 // CHECK1-NEXT: entry:
339 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
340 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
341 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
342 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
343 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
344 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
345 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
346 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
347 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
348 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
349 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
350 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
351 // CHECK1-NEXT: ret void
352 //
353 //
354 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
355 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
356 // CHECK1-NEXT: entry:
357 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
358 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
359 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
360 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
361 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
362 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
363 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
364 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
365 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
366 // CHECK1: omp.arrayinit.body:
367 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
368 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
369 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
370 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
371 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
372 // CHECK1: omp.arrayinit.done:
373 // CHECK1-NEXT: ret void
374 //
375 //
376 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
377 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
378 // CHECK1-NEXT: entry:
379 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
380 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
381 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
382 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
383 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
384 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
385 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
386 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
387 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
388 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
389 // CHECK1: omp.arraycpy.body:
390 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
391 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
392 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
393 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
394 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
395 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
396 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
397 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
398 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
399 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
400 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
401 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
402 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
403 // CHECK1: omp.arraycpy.done4:
404 // CHECK1-NEXT: ret void
405 //
406 //
407 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
408 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1]] {
409 // CHECK1-NEXT: entry:
410 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
411 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
412 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
413 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
414 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
415 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
416 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
417 // CHECK1-NEXT: [[TMP:%.*]] = alloca i64, align 8
418 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
419 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
420 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
421 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
422 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
423 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
424 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
425 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
426 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
427 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t.0], align 8
428 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
429 // CHECK1-NEXT: [[I:%.*]] = alloca i64, align 8
430 // CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
431 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
432 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
433 // CHECK1-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
434 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
435 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
436 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
437 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
438 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
439 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
440 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
441 // CHECK1-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
442 // CHECK1-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
443 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
444 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
445 // CHECK1-NEXT: store i64 [[TMP1]], i64* [[DOTOMP_LB]], align 8
446 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[DOTOMP_UB]], align 8
447 // CHECK1-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
448 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
449 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
450 // CHECK1-NEXT: [[TMP3:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
451 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP3]], i64 0
452 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
453 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP4]], i64 0
454 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
455 // CHECK1-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
456 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP6]]
457 // CHECK1-NEXT: [[TMP7:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
458 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP7]], i64 9
459 // CHECK1-NEXT: [[TMP8:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
460 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP8]], i64 [[LB_ADD_LEN]]
461 // CHECK1-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
462 // CHECK1-NEXT: [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
463 // CHECK1-NEXT: [[TMP11:%.*]] = sub i64 [[TMP9]], [[TMP10]]
464 // CHECK1-NEXT: [[TMP12:%.*]] = sdiv exact i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
465 // CHECK1-NEXT: [[TMP13:%.*]] = add nuw i64 [[TMP12]], 1
466 // CHECK1-NEXT: [[TMP14:%.*]] = mul nuw i64 [[TMP13]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
467 // CHECK1-NEXT: [[TMP15:%.*]] = call i8* @llvm.stacksave()
468 // CHECK1-NEXT: store i8* [[TMP15]], i8** [[SAVED_STACK]], align 8
469 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP13]], align 16
470 // CHECK1-NEXT: store i64 [[TMP13]], i64* [[__VLA_EXPR0]], align 8
471 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP13]]
472 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP16]]
473 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
474 // CHECK1: omp.arrayinit.body:
475 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
476 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
477 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
478 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP16]]
479 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
480 // CHECK1: omp.arrayinit.done:
481 // CHECK1-NEXT: [[TMP17:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
482 // CHECK1-NEXT: [[TMP18:%.*]] = load i8*, i8** [[TMP17]], align 8
483 // CHECK1-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[TMP18]] to i64
484 // CHECK1-NEXT: [[TMP20:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
485 // CHECK1-NEXT: [[TMP21:%.*]] = sub i64 [[TMP19]], [[TMP20]]
486 // CHECK1-NEXT: [[TMP22:%.*]] = sdiv exact i64 [[TMP21]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
487 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP22]]
488 // CHECK1-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
489 // CHECK1-NEXT: store i8* [[TMP23]], i8** [[_TMP6]], align 8
490 // CHECK1-NEXT: [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t.0], [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]], i64 0, i64 0
491 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0:%.*]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
492 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i32* [[ARGC1]] to i8*
493 // CHECK1-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
494 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
495 // CHECK1-NEXT: [[TMP27:%.*]] = bitcast i32* [[TMP0]] to i8*
496 // CHECK1-NEXT: store i8* [[TMP27]], i8** [[TMP26]], align 8
497 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
498 // CHECK1-NEXT: store i64 4, i64* [[TMP28]], align 8
499 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
500 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..4 to i8*), i8** [[TMP29]], align 8
501 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
502 // CHECK1-NEXT: store i8* null, i8** [[TMP30]], align 8
503 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
504 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..5 to i8*), i8** [[TMP31]], align 8
505 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
506 // CHECK1-NEXT: [[TMP33:%.*]] = bitcast i32* [[TMP32]] to i8*
507 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP33]], i8 0, i64 4, i1 false)
508 // CHECK1-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t.0], [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]], i64 0, i64 1
509 // CHECK1-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
510 // CHECK1-NEXT: [[TMP35:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
511 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP35]], i64 0
512 // CHECK1-NEXT: [[TMP36:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
513 // CHECK1-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP36]], i64 0
514 // CHECK1-NEXT: [[TMP37:%.*]] = load i32, i32* [[TMP0]], align 4
515 // CHECK1-NEXT: [[TMP38:%.*]] = sext i32 [[TMP37]] to i64
516 // CHECK1-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP38]]
517 // CHECK1-NEXT: [[TMP39:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
518 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP39]], i64 9
519 // CHECK1-NEXT: [[TMP40:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
520 // CHECK1-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP40]], i64 [[LB_ADD_LEN10]]
521 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP34]], align 8
522 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
523 // CHECK1-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP41]], align 8
524 // CHECK1-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
525 // CHECK1-NEXT: [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
526 // CHECK1-NEXT: [[TMP44:%.*]] = sub i64 [[TMP42]], [[TMP43]]
527 // CHECK1-NEXT: [[TMP45:%.*]] = sdiv exact i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
528 // CHECK1-NEXT: [[TMP46:%.*]] = add nuw i64 [[TMP45]], 1
529 // CHECK1-NEXT: [[TMP47:%.*]] = mul nuw i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
530 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
531 // CHECK1-NEXT: store i64 [[TMP47]], i64* [[TMP48]], align 8
532 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
533 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..6 to i8*), i8** [[TMP49]], align 8
534 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
535 // CHECK1-NEXT: store i8* null, i8** [[TMP50]], align 8
536 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
537 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..7 to i8*), i8** [[TMP51]], align 8
538 // CHECK1-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
539 // CHECK1-NEXT: store i32 1, i32* [[TMP52]], align 8
540 // CHECK1-NEXT: [[TMP53:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
541 // CHECK1-NEXT: [[TMP54:%.*]] = load i32, i32* [[TMP53]], align 4
542 // CHECK1-NEXT: [[TMP55:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]] to i8*
543 // CHECK1-NEXT: [[TMP56:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP54]], i32 1, i32 2, i8* [[TMP55]])
544 // CHECK1-NEXT: store i8* [[TMP56]], i8** [[DOTTASK_RED_]], align 8
545 // CHECK1-NEXT: [[TMP57:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
546 // CHECK1-NEXT: [[TMP58:%.*]] = load i32, i32* [[TMP57]], align 4
547 // CHECK1-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP58]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
548 // CHECK1-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
549 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP59]], 9
550 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
551 // CHECK1: cond.true:
552 // CHECK1-NEXT: br label [[COND_END:%.*]]
553 // CHECK1: cond.false:
554 // CHECK1-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
555 // CHECK1-NEXT: br label [[COND_END]]
556 // CHECK1: cond.end:
557 // CHECK1-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP60]], [[COND_FALSE]] ]
558 // CHECK1-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
559 // CHECK1-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
560 // CHECK1-NEXT: store i64 [[TMP61]], i64* [[DOTOMP_IV]], align 8
561 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
562 // CHECK1: omp.inner.for.cond:
563 // CHECK1-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
564 // CHECK1-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
565 // CHECK1-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP62]], [[TMP63]]
566 // CHECK1-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
567 // CHECK1: omp.inner.for.cond.cleanup:
568 // CHECK1-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
569 // CHECK1: omp.inner.for.body:
570 // CHECK1-NEXT: [[TMP64:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
571 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP64]], 1
572 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
573 // CHECK1-NEXT: store i64 [[ADD]], i64* [[I]], align 8
574 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
575 // CHECK1-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP65]], align 8
576 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
577 // CHECK1-NEXT: store i32* [[ARGC1]], i32** [[TMP66]], align 8
578 // CHECK1-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
579 // CHECK1-NEXT: [[TMP68:%.*]] = load i8**, i8*** [[_TMP5]], align 8
580 // CHECK1-NEXT: store i8** [[TMP68]], i8*** [[TMP67]], align 8
581 // CHECK1-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
582 // CHECK1-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
583 // CHECK1-NEXT: [[TMP71:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP70]], 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*)*))
584 // CHECK1-NEXT: [[TMP72:%.*]] = bitcast i8* [[TMP71]] to %struct.kmp_task_t_with_privates*
585 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 0
586 // CHECK1-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP73]], i32 0, i32 0
587 // CHECK1-NEXT: [[TMP75:%.*]] = load i8*, i8** [[TMP74]], align 8
588 // CHECK1-NEXT: [[TMP76:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
589 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP75]], i8* align 8 [[TMP76]], i64 24, i1 false)
590 // CHECK1-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 1
591 // CHECK1-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP77]], i32 0, i32 0
592 // CHECK1-NEXT: [[TMP79:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
593 // CHECK1-NEXT: store i8* [[TMP79]], i8** [[TMP78]], align 8
594 // CHECK1-NEXT: [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
595 // CHECK1-NEXT: [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
596 // CHECK1-NEXT: [[TMP82:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i8* [[TMP71]])
597 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
598 // CHECK1: omp.body.continue:
599 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
600 // CHECK1: omp.inner.for.inc:
601 // CHECK1-NEXT: [[TMP83:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
602 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP83]], 1
603 // CHECK1-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
604 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
605 // CHECK1: omp.inner.for.end:
606 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
607 // CHECK1: omp.loop.exit:
608 // CHECK1-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
609 // CHECK1-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
610 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]])
611 // CHECK1-NEXT: [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
612 // CHECK1-NEXT: [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
613 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP87]], i32 1)
614 // CHECK1-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
615 // CHECK1-NEXT: [[TMP89:%.*]] = bitcast i32* [[ARGC1]] to i8*
616 // CHECK1-NEXT: store i8* [[TMP89]], i8** [[TMP88]], align 8
617 // CHECK1-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
618 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP90]], align 8
619 // CHECK1-NEXT: [[TMP91:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
620 // CHECK1-NEXT: [[TMP92:%.*]] = inttoptr i64 [[TMP13]] to i8*
621 // CHECK1-NEXT: store i8* [[TMP92]], i8** [[TMP91]], align 8
622 // CHECK1-NEXT: [[TMP93:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
623 // CHECK1-NEXT: [[TMP94:%.*]] = load i32, i32* [[TMP93]], align 4
624 // CHECK1-NEXT: [[TMP95:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
625 // CHECK1-NEXT: [[TMP96:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP94]], i32 2, i64 24, i8* [[TMP95]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
626 // CHECK1-NEXT: switch i32 [[TMP96]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
627 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
628 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
629 // CHECK1-NEXT: ]
630 // CHECK1: .omp.reduction.case1:
631 // CHECK1-NEXT: [[TMP97:%.*]] = load i32, i32* [[TMP0]], align 4
632 // CHECK1-NEXT: [[TMP98:%.*]] = load i32, i32* [[ARGC1]], align 4
633 // CHECK1-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP97]], [[TMP98]]
634 // CHECK1-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
635 // CHECK1-NEXT: [[TMP99:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP13]]
636 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP99]]
637 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
638 // CHECK1: omp.arraycpy.body:
639 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
640 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
641 // CHECK1-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
642 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP100]] to i32
643 // CHECK1-NEXT: [[TMP101:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
644 // CHECK1-NEXT: [[CONV17:%.*]] = sext i8 [[TMP101]] to i32
645 // CHECK1-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
646 // CHECK1-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
647 // CHECK1-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
648 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
649 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
650 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP99]]
651 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
652 // CHECK1: omp.arraycpy.done22:
653 // CHECK1-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
654 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
655 // CHECK1: .omp.reduction.case2:
656 // CHECK1-NEXT: [[TMP102:%.*]] = load i32, i32* [[ARGC1]], align 4
657 // CHECK1-NEXT: [[TMP103:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP102]] monotonic, align 4
658 // CHECK1-NEXT: [[TMP104:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP13]]
659 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP104]]
660 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
661 // CHECK1: omp.arraycpy.body24:
662 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
663 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
664 // CHECK1-NEXT: [[TMP105:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
665 // CHECK1-NEXT: [[CONV27:%.*]] = sext i8 [[TMP105]] to i32
666 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
667 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
668 // CHECK1: atomic_cont:
669 // CHECK1-NEXT: [[TMP106:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP111:%.*]], [[ATOMIC_CONT]] ]
670 // CHECK1-NEXT: store i8 [[TMP106]], i8* [[_TMP28]], align 1
671 // CHECK1-NEXT: [[TMP107:%.*]] = load i8, i8* [[_TMP28]], align 1
672 // CHECK1-NEXT: [[CONV29:%.*]] = sext i8 [[TMP107]] to i32
673 // CHECK1-NEXT: [[TMP108:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
674 // CHECK1-NEXT: [[CONV30:%.*]] = sext i8 [[TMP108]] to i32
675 // CHECK1-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
676 // CHECK1-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
677 // CHECK1-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
678 // CHECK1-NEXT: [[TMP109:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
679 // CHECK1-NEXT: [[TMP110:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP106]], i8 [[TMP109]] monotonic monotonic, align 1
680 // CHECK1-NEXT: [[TMP111]] = extractvalue { i8, i1 } [[TMP110]], 0
681 // CHECK1-NEXT: [[TMP112:%.*]] = extractvalue { i8, i1 } [[TMP110]], 1
682 // CHECK1-NEXT: br i1 [[TMP112]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
683 // CHECK1: atomic_exit:
684 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
685 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
686 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP104]]
687 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
688 // CHECK1: omp.arraycpy.done36:
689 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
690 // CHECK1: .omp.reduction.default:
691 // CHECK1-NEXT: [[TMP113:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
692 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP113]])
693 // CHECK1-NEXT: ret void
694 //
695 //
696 // CHECK1-LABEL: define {{[^@]+}}@.red_init..4
697 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
698 // CHECK1-NEXT: entry:
699 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
700 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
701 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
702 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
703 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
704 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
705 // CHECK1-NEXT: store i32 0, i32* [[TMP3]], align 8
706 // CHECK1-NEXT: ret void
707 //
708 //
709 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..5
710 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
711 // CHECK1-NEXT: entry:
712 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
713 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
714 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
715 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
716 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
717 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
718 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
719 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
720 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
721 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
722 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
723 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
724 // CHECK1-NEXT: ret void
725 //
726 //
727 // CHECK1-LABEL: define {{[^@]+}}@.red_init..6
728 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
729 // CHECK1-NEXT: entry:
730 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
731 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
732 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
733 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
734 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
735 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
736 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
737 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
738 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
739 // CHECK1: omp.arrayinit.body:
740 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
741 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
742 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
743 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
744 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
745 // CHECK1: omp.arrayinit.done:
746 // CHECK1-NEXT: ret void
747 //
748 //
749 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..7
750 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
751 // CHECK1-NEXT: entry:
752 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
753 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
754 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
755 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
756 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
757 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
758 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
759 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
760 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
761 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
762 // CHECK1: omp.arraycpy.body:
763 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
764 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
765 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
766 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
767 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
768 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
769 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
770 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
771 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
772 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
773 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
774 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
775 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
776 // CHECK1: omp.arraycpy.done4:
777 // CHECK1-NEXT: ret void
778 //
779 //
780 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
781 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
782 // CHECK1-NEXT: entry:
783 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
784 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
785 // CHECK1-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
786 // CHECK1-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
787 // CHECK1-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
788 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
789 // CHECK1-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
790 // CHECK1-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
791 // CHECK1-NEXT: ret void
792 //
793 //
794 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
795 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
796 // CHECK1-NEXT: entry:
797 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
798 // CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
799 // CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
800 // CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
801 // CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
802 // CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
803 // CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
804 // CHECK1-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
805 // CHECK1-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
806 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
807 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
808 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
809 // CHECK1-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
810 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
811 // CHECK1-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
812 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
813 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
814 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
815 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
816 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
817 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
818 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
819 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
820 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
821 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
822 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
823 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
824 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
825 // CHECK1-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
826 // CHECK1-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
827 // 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
828 // CHECK1-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
829 // CHECK1-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
830 // CHECK1-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
831 // CHECK1-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
832 // CHECK1-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
833 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
834 // CHECK1-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5]]
835 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
836 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
837 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
838 // CHECK1-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
839 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
840 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
841 // CHECK1-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
842 // CHECK1-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
843 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
844 // CHECK1-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
845 // CHECK1-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
846 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
847 // CHECK1-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
848 // CHECK1-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
849 // CHECK1-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
850 // CHECK1-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
851 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
852 // CHECK1-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
853 // CHECK1-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
854 // CHECK1-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
855 // CHECK1-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
856 // CHECK1-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
857 // CHECK1-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
858 // CHECK1-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
859 // CHECK1-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
860 // CHECK1-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
861 // CHECK1-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
862 // CHECK1-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
863 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
864 // CHECK1-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
865 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
866 // CHECK1-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
867 // CHECK1-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
868 // CHECK1-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
869 // CHECK1-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
870 // CHECK1-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
871 // CHECK1-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
872 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
873 // CHECK1-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
874 // CHECK1-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
875 // CHECK1-NEXT: ret i32 0
876 //
877 //
878 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
879 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
880 // CHECK1-NEXT: entry:
881 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
882 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
883 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
884 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
885 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
886 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
887 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
888 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
889 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
890 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
891 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
892 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
893 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
894 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
895 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
896 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
897 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
898 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
899 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
900 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
901 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
902 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
903 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
904 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
905 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
906 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
907 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
908 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
909 // CHECK1: omp.arraycpy.body:
910 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
911 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
912 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
913 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
914 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
915 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
916 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
917 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
918 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
919 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
920 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
921 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
922 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
923 // CHECK1: omp.arraycpy.done5:
924 // CHECK1-NEXT: ret void
925 //
926 //
927 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.9
928 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
929 // CHECK1-NEXT: entry:
930 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
931 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
932 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
933 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
934 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
935 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
936 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
937 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
938 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
939 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
940 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
941 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
942 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
943 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
944 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
945 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
946 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
947 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
948 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
949 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
950 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
951 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
952 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
953 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
954 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
955 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
956 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
957 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
958 // CHECK1: omp.arraycpy.body:
959 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
960 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
961 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
962 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
963 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
964 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
965 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
966 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
967 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
968 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
969 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
970 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
971 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
972 // CHECK1: omp.arraycpy.done5:
973 // CHECK1-NEXT: ret void
974 //
975 //
976 // CHECK2-LABEL: define {{[^@]+}}@main
977 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
978 // CHECK2-NEXT: entry:
979 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
980 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
981 // CHECK2-NEXT: [[ARGC_CASTED:%.*]] = alloca i64, align 8
982 // CHECK2-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
983 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
984 // CHECK2-NEXT: [[TMP0:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
985 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_CASTED]] to i32*
986 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
987 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[ARGC_CASTED]], align 8
988 // CHECK2-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
989 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14(i64 [[TMP1]], i8** [[TMP2]]) #[[ATTR5:[0-9]+]]
990 // CHECK2-NEXT: ret i32 0
991 //
992 //
993 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l14
994 // CHECK2-SAME: (i64 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
995 // CHECK2-NEXT: entry:
996 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i64, align 8
997 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
998 // CHECK2-NEXT: store i64 [[ARGC]], i64* [[ARGC_ADDR]], align 8
999 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1000 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[ARGC_ADDR]] to i32*
1001 // CHECK2-NEXT: [[TMP0:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1002 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8**)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i8** [[TMP0]])
1003 // CHECK2-NEXT: ret void
1004 //
1005 //
1006 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1007 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1]] {
1008 // CHECK2-NEXT: entry:
1009 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1010 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1011 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1012 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1013 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
1014 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1015 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1016 // CHECK2-NEXT: [[TMP:%.*]] = alloca i8**, align 8
1017 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
1018 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
1019 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
1020 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
1021 // CHECK2-NEXT: [[_TMP12:%.*]] = alloca i64, align 8
1022 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i64, align 8
1023 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i64, align 8
1024 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1025 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1026 // CHECK2-NEXT: [[I:%.*]] = alloca i64, align 8
1027 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
1028 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
1029 // CHECK2-NEXT: [[_TMP27:%.*]] = alloca i8, align 1
1030 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1031 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1032 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1033 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1034 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1035 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
1036 // CHECK2-NEXT: [[TMP1:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1037 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP1]], i64 0
1038 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
1039 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP2]], i64 0
1040 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1041 // CHECK2-NEXT: [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
1042 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP4]]
1043 // CHECK2-NEXT: [[TMP5:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1044 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP5]], i64 9
1045 // CHECK2-NEXT: [[TMP6:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
1046 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP6]], i64 [[LB_ADD_LEN]]
1047 // CHECK2-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
1048 // CHECK2-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
1049 // CHECK2-NEXT: [[TMP9:%.*]] = sub i64 [[TMP7]], [[TMP8]]
1050 // CHECK2-NEXT: [[TMP10:%.*]] = sdiv exact i64 [[TMP9]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1051 // CHECK2-NEXT: [[TMP11:%.*]] = add nuw i64 [[TMP10]], 1
1052 // CHECK2-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1053 // CHECK2-NEXT: [[TMP13:%.*]] = call i8* @llvm.stacksave()
1054 // CHECK2-NEXT: store i8* [[TMP13]], i8** [[SAVED_STACK]], align 8
1055 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP11]], align 16
1056 // CHECK2-NEXT: store i64 [[TMP11]], i64* [[__VLA_EXPR0]], align 8
1057 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP11]]
1058 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP14]]
1059 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
1060 // CHECK2: omp.arrayinit.body:
1061 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
1062 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1063 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1064 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP14]]
1065 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
1066 // CHECK2: omp.arrayinit.done:
1067 // CHECK2-NEXT: [[TMP15:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1068 // CHECK2-NEXT: [[TMP16:%.*]] = load i8*, i8** [[TMP15]], align 8
1069 // CHECK2-NEXT: [[TMP17:%.*]] = ptrtoint i8* [[TMP16]] to i64
1070 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
1071 // CHECK2-NEXT: [[TMP19:%.*]] = sub i64 [[TMP17]], [[TMP18]]
1072 // CHECK2-NEXT: [[TMP20:%.*]] = sdiv exact i64 [[TMP19]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1073 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP20]]
1074 // CHECK2-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
1075 // CHECK2-NEXT: store i8* [[TMP21]], i8** [[_TMP5]], align 8
1076 // 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
1077 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
1078 // CHECK2-NEXT: [[TMP23:%.*]] = bitcast i32* [[ARGC1]] to i8*
1079 // CHECK2-NEXT: store i8* [[TMP23]], i8** [[TMP22]], align 8
1080 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
1081 // CHECK2-NEXT: [[TMP25:%.*]] = bitcast i32* [[TMP0]] to i8*
1082 // CHECK2-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
1083 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
1084 // CHECK2-NEXT: store i64 4, i64* [[TMP26]], align 8
1085 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
1086 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP27]], align 8
1087 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
1088 // CHECK2-NEXT: store i8* null, i8** [[TMP28]], align 8
1089 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
1090 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP29]], align 8
1091 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
1092 // CHECK2-NEXT: [[TMP31:%.*]] = bitcast i32* [[TMP30]] to i8*
1093 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP31]], i8 0, i64 4, i1 false)
1094 // 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
1095 // CHECK2-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
1096 // CHECK2-NEXT: [[TMP33:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1097 // CHECK2-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP33]], i64 0
1098 // CHECK2-NEXT: [[TMP34:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
1099 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP34]], i64 0
1100 // CHECK2-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP0]], align 4
1101 // CHECK2-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
1102 // CHECK2-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP36]]
1103 // CHECK2-NEXT: [[TMP37:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1104 // CHECK2-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP37]], i64 9
1105 // CHECK2-NEXT: [[TMP38:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
1106 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP38]], i64 [[LB_ADD_LEN9]]
1107 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP32]], align 8
1108 // CHECK2-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
1109 // CHECK2-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP39]], align 8
1110 // CHECK2-NEXT: [[TMP40:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
1111 // CHECK2-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
1112 // CHECK2-NEXT: [[TMP42:%.*]] = sub i64 [[TMP40]], [[TMP41]]
1113 // CHECK2-NEXT: [[TMP43:%.*]] = sdiv exact i64 [[TMP42]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1114 // CHECK2-NEXT: [[TMP44:%.*]] = add nuw i64 [[TMP43]], 1
1115 // CHECK2-NEXT: [[TMP45:%.*]] = mul nuw i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1116 // CHECK2-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
1117 // CHECK2-NEXT: store i64 [[TMP45]], i64* [[TMP46]], align 8
1118 // CHECK2-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
1119 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP47]], align 8
1120 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
1121 // CHECK2-NEXT: store i8* null, i8** [[TMP48]], align 8
1122 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
1123 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP49]], align 8
1124 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
1125 // CHECK2-NEXT: store i32 1, i32* [[TMP50]], align 8
1126 // CHECK2-NEXT: [[TMP51:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1127 // CHECK2-NEXT: [[TMP52:%.*]] = load i32, i32* [[TMP51]], align 4
1128 // CHECK2-NEXT: [[TMP53:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
1129 // CHECK2-NEXT: [[TMP54:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP52]], i32 1, i32 2, i8* [[TMP53]])
1130 // CHECK2-NEXT: store i8* [[TMP54]], i8** [[DOTTASK_RED_]], align 8
1131 // CHECK2-NEXT: store i64 0, i64* [[DOTOMP_COMB_LB]], align 8
1132 // CHECK2-NEXT: store i64 9, i64* [[DOTOMP_COMB_UB]], align 8
1133 // CHECK2-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1134 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1135 // CHECK2-NEXT: [[TMP55:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1136 // CHECK2-NEXT: [[TMP56:%.*]] = load i32, i32* [[TMP55]], align 4
1137 // CHECK2-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP56]], i32 92, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_COMB_LB]], i64* [[DOTOMP_COMB_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
1138 // CHECK2-NEXT: [[TMP57:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
1139 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP57]], 9
1140 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1141 // CHECK2: cond.true:
1142 // CHECK2-NEXT: br label [[COND_END:%.*]]
1143 // CHECK2: cond.false:
1144 // CHECK2-NEXT: [[TMP58:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
1145 // CHECK2-NEXT: br label [[COND_END]]
1146 // CHECK2: cond.end:
1147 // CHECK2-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP58]], [[COND_FALSE]] ]
1148 // CHECK2-NEXT: store i64 [[COND]], i64* [[DOTOMP_COMB_UB]], align 8
1149 // CHECK2-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
1150 // CHECK2-NEXT: store i64 [[TMP59]], i64* [[DOTOMP_IV]], align 8
1151 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1152 // CHECK2: omp.inner.for.cond:
1153 // CHECK2-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1154 // CHECK2-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
1155 // CHECK2-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP60]], [[TMP61]]
1156 // CHECK2-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1157 // CHECK2: omp.inner.for.body:
1158 // CHECK2-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_COMB_LB]], align 8
1159 // CHECK2-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_COMB_UB]], align 8
1160 // CHECK2-NEXT: [[TMP64:%.*]] = load i8**, i8*** [[TMP]], align 8
1161 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i8**)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP62]], i64 [[TMP63]], i32* [[ARGC1]], i8** [[TMP64]])
1162 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1163 // CHECK2: omp.inner.for.inc:
1164 // CHECK2-NEXT: [[TMP65:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1165 // CHECK2-NEXT: [[TMP66:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
1166 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i64 [[TMP65]], [[TMP66]]
1167 // CHECK2-NEXT: store i64 [[ADD]], i64* [[DOTOMP_IV]], align 8
1168 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1169 // CHECK2: omp.inner.for.end:
1170 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1171 // CHECK2: omp.loop.exit:
1172 // CHECK2-NEXT: [[TMP67:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1173 // CHECK2-NEXT: [[TMP68:%.*]] = load i32, i32* [[TMP67]], align 4
1174 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP68]])
1175 // CHECK2-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1176 // CHECK2-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
1177 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP70]], i32 1)
1178 // CHECK2-NEXT: [[TMP71:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1179 // CHECK2-NEXT: [[TMP72:%.*]] = bitcast i32* [[ARGC1]] to i8*
1180 // CHECK2-NEXT: store i8* [[TMP72]], i8** [[TMP71]], align 8
1181 // CHECK2-NEXT: [[TMP73:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
1182 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP73]], align 8
1183 // CHECK2-NEXT: [[TMP74:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
1184 // CHECK2-NEXT: [[TMP75:%.*]] = inttoptr i64 [[TMP11]] to i8*
1185 // CHECK2-NEXT: store i8* [[TMP75]], i8** [[TMP74]], align 8
1186 // CHECK2-NEXT: [[TMP76:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1187 // CHECK2-NEXT: [[TMP77:%.*]] = load i32, i32* [[TMP76]], align 4
1188 // CHECK2-NEXT: [[TMP78:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1189 // CHECK2-NEXT: [[TMP79:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP77]], i32 2, i64 24, i8* [[TMP78]], void (i8*, i8*)* @.omp.reduction.reduction_func.9, [8 x i32]* @.gomp_critical_user_.reduction.var)
1190 // CHECK2-NEXT: switch i32 [[TMP79]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1191 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1192 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1193 // CHECK2-NEXT: ]
1194 // CHECK2: .omp.reduction.case1:
1195 // CHECK2-NEXT: [[TMP80:%.*]] = load i32, i32* [[TMP0]], align 4
1196 // CHECK2-NEXT: [[TMP81:%.*]] = load i32, i32* [[ARGC1]], align 4
1197 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP80]], [[TMP81]]
1198 // CHECK2-NEXT: store i32 [[ADD14]], i32* [[TMP0]], align 4
1199 // CHECK2-NEXT: [[TMP82:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
1200 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP82]]
1201 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE21:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1202 // CHECK2: omp.arraycpy.body:
1203 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1204 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST15:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT19:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1205 // CHECK2-NEXT: [[TMP83:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 1
1206 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP83]] to i32
1207 // CHECK2-NEXT: [[TMP84:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1208 // CHECK2-NEXT: [[CONV16:%.*]] = sext i8 [[TMP84]] to i32
1209 // CHECK2-NEXT: [[ADD17:%.*]] = add nsw i32 [[CONV]], [[CONV16]]
1210 // CHECK2-NEXT: [[CONV18:%.*]] = trunc i32 [[ADD17]] to i8
1211 // CHECK2-NEXT: store i8 [[CONV18]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], align 1
1212 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT19]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST15]], i32 1
1213 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1214 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE20:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT19]], [[TMP82]]
1215 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE20]], label [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_BODY]]
1216 // CHECK2: omp.arraycpy.done21:
1217 // CHECK2-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP77]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1218 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1219 // CHECK2: .omp.reduction.case2:
1220 // CHECK2-NEXT: [[TMP85:%.*]] = load i32, i32* [[ARGC1]], align 4
1221 // CHECK2-NEXT: [[TMP86:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP85]] monotonic, align 4
1222 // CHECK2-NEXT: [[TMP87:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP11]]
1223 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY22:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP87]]
1224 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY22]], label [[OMP_ARRAYCPY_DONE35:%.*]], label [[OMP_ARRAYCPY_BODY23:%.*]]
1225 // CHECK2: omp.arraycpy.body23:
1226 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST24:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT33:%.*]], [[ATOMIC_EXIT:%.*]] ]
1227 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST25:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT32:%.*]], [[ATOMIC_EXIT]] ]
1228 // CHECK2-NEXT: [[TMP88:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], align 1
1229 // CHECK2-NEXT: [[CONV26:%.*]] = sext i8 [[TMP88]] to i32
1230 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]] monotonic, align 1
1231 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
1232 // CHECK2: atomic_cont:
1233 // CHECK2-NEXT: [[TMP89:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY23]] ], [ [[TMP94:%.*]], [[ATOMIC_CONT]] ]
1234 // CHECK2-NEXT: store i8 [[TMP89]], i8* [[_TMP27]], align 1
1235 // CHECK2-NEXT: [[TMP90:%.*]] = load i8, i8* [[_TMP27]], align 1
1236 // CHECK2-NEXT: [[CONV28:%.*]] = sext i8 [[TMP90]] to i32
1237 // CHECK2-NEXT: [[TMP91:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], align 1
1238 // CHECK2-NEXT: [[CONV29:%.*]] = sext i8 [[TMP91]] to i32
1239 // CHECK2-NEXT: [[ADD30:%.*]] = add nsw i32 [[CONV28]], [[CONV29]]
1240 // CHECK2-NEXT: [[CONV31:%.*]] = trunc i32 [[ADD30]] to i8
1241 // CHECK2-NEXT: store i8 [[CONV31]], i8* [[ATOMIC_TEMP]], align 1
1242 // CHECK2-NEXT: [[TMP92:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
1243 // CHECK2-NEXT: [[TMP93:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]], i8 [[TMP89]], i8 [[TMP92]] monotonic monotonic, align 1
1244 // CHECK2-NEXT: [[TMP94]] = extractvalue { i8, i1 } [[TMP93]], 0
1245 // CHECK2-NEXT: [[TMP95:%.*]] = extractvalue { i8, i1 } [[TMP93]], 1
1246 // CHECK2-NEXT: br i1 [[TMP95]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
1247 // CHECK2: atomic_exit:
1248 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT32]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST25]], i32 1
1249 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST24]], i32 1
1250 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE34:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT32]], [[TMP87]]
1251 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE34]], label [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_BODY23]]
1252 // CHECK2: omp.arraycpy.done35:
1253 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1254 // CHECK2: .omp.reduction.default:
1255 // CHECK2-NEXT: [[TMP96:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1256 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP96]])
1257 // CHECK2-NEXT: ret void
1258 //
1259 //
1260 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
1261 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
1262 // CHECK2-NEXT: entry:
1263 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1264 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1265 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1266 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1267 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1268 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
1269 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
1270 // CHECK2-NEXT: ret void
1271 //
1272 //
1273 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
1274 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1275 // CHECK2-NEXT: entry:
1276 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1277 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1278 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1279 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1280 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1281 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
1282 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1283 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
1284 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
1285 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
1286 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
1287 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
1288 // CHECK2-NEXT: ret void
1289 //
1290 //
1291 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
1292 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
1293 // CHECK2-NEXT: entry:
1294 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1295 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1296 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1297 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1298 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1299 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
1300 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
1301 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
1302 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
1303 // CHECK2: omp.arrayinit.body:
1304 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
1305 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1306 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1307 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
1308 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
1309 // CHECK2: omp.arrayinit.done:
1310 // CHECK2-NEXT: ret void
1311 //
1312 //
1313 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
1314 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1315 // CHECK2-NEXT: entry:
1316 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1317 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1318 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1319 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1320 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
1321 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1322 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1323 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
1324 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
1325 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1326 // CHECK2: omp.arraycpy.body:
1327 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1328 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1329 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1330 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
1331 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1332 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
1333 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1334 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
1335 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1336 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1337 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1338 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
1339 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
1340 // CHECK2: omp.arraycpy.done4:
1341 // CHECK2-NEXT: ret void
1342 //
1343 //
1344 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1345 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR1]] {
1346 // CHECK2-NEXT: entry:
1347 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1348 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1349 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1350 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1351 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
1352 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1353 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i64, align 8
1354 // CHECK2-NEXT: [[TMP:%.*]] = alloca i64, align 8
1355 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i64, align 8
1356 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i64, align 8
1357 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1358 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1359 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
1360 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1361 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1362 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8**, align 8
1363 // CHECK2-NEXT: [[_TMP6:%.*]] = alloca i8*, align 8
1364 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t.0], align 8
1365 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
1366 // CHECK2-NEXT: [[I:%.*]] = alloca i64, align 8
1367 // CHECK2-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
1368 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
1369 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
1370 // CHECK2-NEXT: [[_TMP28:%.*]] = alloca i8, align 1
1371 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1372 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1373 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1374 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1375 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
1376 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1377 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
1378 // CHECK2-NEXT: store i64 0, i64* [[DOTOMP_LB]], align 8
1379 // CHECK2-NEXT: store i64 9, i64* [[DOTOMP_UB]], align 8
1380 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1381 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1382 // CHECK2-NEXT: store i64 [[TMP1]], i64* [[DOTOMP_LB]], align 8
1383 // CHECK2-NEXT: store i64 [[TMP2]], i64* [[DOTOMP_UB]], align 8
1384 // CHECK2-NEXT: store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1385 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1386 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
1387 // CHECK2-NEXT: [[TMP3:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1388 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP3]], i64 0
1389 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
1390 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP4]], i64 0
1391 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP0]], align 4
1392 // CHECK2-NEXT: [[TMP6:%.*]] = sext i32 [[TMP5]] to i64
1393 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP6]]
1394 // CHECK2-NEXT: [[TMP7:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1395 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP7]], i64 9
1396 // CHECK2-NEXT: [[TMP8:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
1397 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP8]], i64 [[LB_ADD_LEN]]
1398 // CHECK2-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
1399 // CHECK2-NEXT: [[TMP10:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
1400 // CHECK2-NEXT: [[TMP11:%.*]] = sub i64 [[TMP9]], [[TMP10]]
1401 // CHECK2-NEXT: [[TMP12:%.*]] = sdiv exact i64 [[TMP11]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1402 // CHECK2-NEXT: [[TMP13:%.*]] = add nuw i64 [[TMP12]], 1
1403 // CHECK2-NEXT: [[TMP14:%.*]] = mul nuw i64 [[TMP13]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1404 // CHECK2-NEXT: [[TMP15:%.*]] = call i8* @llvm.stacksave()
1405 // CHECK2-NEXT: store i8* [[TMP15]], i8** [[SAVED_STACK]], align 8
1406 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP13]], align 16
1407 // CHECK2-NEXT: store i64 [[TMP13]], i64* [[__VLA_EXPR0]], align 8
1408 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP13]]
1409 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP16]]
1410 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
1411 // CHECK2: omp.arrayinit.body:
1412 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
1413 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1414 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1415 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP16]]
1416 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
1417 // CHECK2: omp.arrayinit.done:
1418 // CHECK2-NEXT: [[TMP17:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1419 // CHECK2-NEXT: [[TMP18:%.*]] = load i8*, i8** [[TMP17]], align 8
1420 // CHECK2-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[TMP18]] to i64
1421 // CHECK2-NEXT: [[TMP20:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
1422 // CHECK2-NEXT: [[TMP21:%.*]] = sub i64 [[TMP19]], [[TMP20]]
1423 // CHECK2-NEXT: [[TMP22:%.*]] = sdiv exact i64 [[TMP21]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1424 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP22]]
1425 // CHECK2-NEXT: store i8** [[_TMP6]], i8*** [[_TMP5]], align 8
1426 // CHECK2-NEXT: store i8* [[TMP23]], i8** [[_TMP6]], align 8
1427 // CHECK2-NEXT: [[DOTRD_INPUT_GEP_:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t.0], [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]], i64 0, i64 0
1428 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0:%.*]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
1429 // CHECK2-NEXT: [[TMP25:%.*]] = bitcast i32* [[ARGC1]] to i8*
1430 // CHECK2-NEXT: store i8* [[TMP25]], i8** [[TMP24]], align 8
1431 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
1432 // CHECK2-NEXT: [[TMP27:%.*]] = bitcast i32* [[TMP0]] to i8*
1433 // CHECK2-NEXT: store i8* [[TMP27]], i8** [[TMP26]], align 8
1434 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
1435 // CHECK2-NEXT: store i64 4, i64* [[TMP28]], align 8
1436 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
1437 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..4 to i8*), i8** [[TMP29]], align 8
1438 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
1439 // CHECK2-NEXT: store i8* null, i8** [[TMP30]], align 8
1440 // CHECK2-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
1441 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..5 to i8*), i8** [[TMP31]], align 8
1442 // CHECK2-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
1443 // CHECK2-NEXT: [[TMP33:%.*]] = bitcast i32* [[TMP32]] to i8*
1444 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP33]], i8 0, i64 4, i1 false)
1445 // CHECK2-NEXT: [[DOTRD_INPUT_GEP_7:%.*]] = getelementptr inbounds [2 x %struct.kmp_taskred_input_t.0], [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]], i64 0, i64 1
1446 // CHECK2-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 0
1447 // CHECK2-NEXT: [[TMP35:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1448 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8*, i8** [[TMP35]], i64 0
1449 // CHECK2-NEXT: [[TMP36:%.*]] = load i8*, i8** [[ARRAYIDX8]], align 8
1450 // CHECK2-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds i8, i8* [[TMP36]], i64 0
1451 // CHECK2-NEXT: [[TMP37:%.*]] = load i32, i32* [[TMP0]], align 4
1452 // CHECK2-NEXT: [[TMP38:%.*]] = sext i32 [[TMP37]] to i64
1453 // CHECK2-NEXT: [[LB_ADD_LEN10:%.*]] = add nsw i64 -1, [[TMP38]]
1454 // CHECK2-NEXT: [[TMP39:%.*]] = load i8**, i8*** [[ARGV_ADDR]], align 8
1455 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8*, i8** [[TMP39]], i64 9
1456 // CHECK2-NEXT: [[TMP40:%.*]] = load i8*, i8** [[ARRAYIDX11]], align 8
1457 // CHECK2-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i8, i8* [[TMP40]], i64 [[LB_ADD_LEN10]]
1458 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP34]], align 8
1459 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
1460 // CHECK2-NEXT: store i8* [[ARRAYIDX9]], i8** [[TMP41]], align 8
1461 // CHECK2-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX12]] to i64
1462 // CHECK2-NEXT: [[TMP43:%.*]] = ptrtoint i8* [[ARRAYIDX9]] to i64
1463 // CHECK2-NEXT: [[TMP44:%.*]] = sub i64 [[TMP42]], [[TMP43]]
1464 // CHECK2-NEXT: [[TMP45:%.*]] = sdiv exact i64 [[TMP44]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1465 // CHECK2-NEXT: [[TMP46:%.*]] = add nuw i64 [[TMP45]], 1
1466 // CHECK2-NEXT: [[TMP47:%.*]] = mul nuw i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1467 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
1468 // CHECK2-NEXT: store i64 [[TMP47]], i64* [[TMP48]], align 8
1469 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
1470 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..6 to i8*), i8** [[TMP49]], align 8
1471 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
1472 // CHECK2-NEXT: store i8* null, i8** [[TMP50]], align 8
1473 // CHECK2-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
1474 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..7 to i8*), i8** [[TMP51]], align 8
1475 // CHECK2-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T_0]], %struct.kmp_taskred_input_t.0* [[DOTRD_INPUT_GEP_7]], i32 0, i32 6
1476 // CHECK2-NEXT: store i32 1, i32* [[TMP52]], align 8
1477 // CHECK2-NEXT: [[TMP53:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1478 // CHECK2-NEXT: [[TMP54:%.*]] = load i32, i32* [[TMP53]], align 4
1479 // CHECK2-NEXT: [[TMP55:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t.0]* [[DOTRD_INPUT_]] to i8*
1480 // CHECK2-NEXT: [[TMP56:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP54]], i32 1, i32 2, i8* [[TMP55]])
1481 // CHECK2-NEXT: store i8* [[TMP56]], i8** [[DOTTASK_RED_]], align 8
1482 // CHECK2-NEXT: [[TMP57:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1483 // CHECK2-NEXT: [[TMP58:%.*]] = load i32, i32* [[TMP57]], align 4
1484 // CHECK2-NEXT: call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP58]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
1485 // CHECK2-NEXT: [[TMP59:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1486 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i64 [[TMP59]], 9
1487 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1488 // CHECK2: cond.true:
1489 // CHECK2-NEXT: br label [[COND_END:%.*]]
1490 // CHECK2: cond.false:
1491 // CHECK2-NEXT: [[TMP60:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1492 // CHECK2-NEXT: br label [[COND_END]]
1493 // CHECK2: cond.end:
1494 // CHECK2-NEXT: [[COND:%.*]] = phi i64 [ 9, [[COND_TRUE]] ], [ [[TMP60]], [[COND_FALSE]] ]
1495 // CHECK2-NEXT: store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
1496 // CHECK2-NEXT: [[TMP61:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1497 // CHECK2-NEXT: store i64 [[TMP61]], i64* [[DOTOMP_IV]], align 8
1498 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1499 // CHECK2: omp.inner.for.cond:
1500 // CHECK2-NEXT: [[TMP62:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1501 // CHECK2-NEXT: [[TMP63:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1502 // CHECK2-NEXT: [[CMP13:%.*]] = icmp sle i64 [[TMP62]], [[TMP63]]
1503 // CHECK2-NEXT: br i1 [[CMP13]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1504 // CHECK2: omp.inner.for.cond.cleanup:
1505 // CHECK2-NEXT: br label [[OMP_INNER_FOR_END:%.*]]
1506 // CHECK2: omp.inner.for.body:
1507 // CHECK2-NEXT: [[TMP64:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1508 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i64 [[TMP64]], 1
1509 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i64 0, [[MUL]]
1510 // CHECK2-NEXT: store i64 [[ADD]], i64* [[I]], align 8
1511 // CHECK2-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
1512 // CHECK2-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP65]], align 8
1513 // CHECK2-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
1514 // CHECK2-NEXT: store i32* [[ARGC1]], i32** [[TMP66]], align 8
1515 // CHECK2-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
1516 // CHECK2-NEXT: [[TMP68:%.*]] = load i8**, i8*** [[_TMP5]], align 8
1517 // CHECK2-NEXT: store i8** [[TMP68]], i8*** [[TMP67]], align 8
1518 // CHECK2-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1519 // CHECK2-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
1520 // CHECK2-NEXT: [[TMP71:%.*]] = call i8* @__kmpc_omp_task_alloc(%struct.ident_t* @[[GLOB1]], i32 [[TMP70]], 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*)*))
1521 // CHECK2-NEXT: [[TMP72:%.*]] = bitcast i8* [[TMP71]] to %struct.kmp_task_t_with_privates*
1522 // CHECK2-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 0
1523 // CHECK2-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP73]], i32 0, i32 0
1524 // CHECK2-NEXT: [[TMP75:%.*]] = load i8*, i8** [[TMP74]], align 8
1525 // CHECK2-NEXT: [[TMP76:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
1526 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP75]], i8* align 8 [[TMP76]], i64 24, i1 false)
1527 // CHECK2-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 1
1528 // CHECK2-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP77]], i32 0, i32 0
1529 // CHECK2-NEXT: [[TMP79:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
1530 // CHECK2-NEXT: store i8* [[TMP79]], i8** [[TMP78]], align 8
1531 // CHECK2-NEXT: [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1532 // CHECK2-NEXT: [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
1533 // CHECK2-NEXT: [[TMP82:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i8* [[TMP71]])
1534 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1535 // CHECK2: omp.body.continue:
1536 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1537 // CHECK2: omp.inner.for.inc:
1538 // CHECK2-NEXT: [[TMP83:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1539 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i64 [[TMP83]], 1
1540 // CHECK2-NEXT: store i64 [[ADD14]], i64* [[DOTOMP_IV]], align 8
1541 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1542 // CHECK2: omp.inner.for.end:
1543 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1544 // CHECK2: omp.loop.exit:
1545 // CHECK2-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1546 // CHECK2-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
1547 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]])
1548 // CHECK2-NEXT: [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1549 // CHECK2-NEXT: [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
1550 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP87]], i32 1)
1551 // CHECK2-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
1552 // CHECK2-NEXT: [[TMP89:%.*]] = bitcast i32* [[ARGC1]] to i8*
1553 // CHECK2-NEXT: store i8* [[TMP89]], i8** [[TMP88]], align 8
1554 // CHECK2-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
1555 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP90]], align 8
1556 // CHECK2-NEXT: [[TMP91:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
1557 // CHECK2-NEXT: [[TMP92:%.*]] = inttoptr i64 [[TMP13]] to i8*
1558 // CHECK2-NEXT: store i8* [[TMP92]], i8** [[TMP91]], align 8
1559 // CHECK2-NEXT: [[TMP93:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1560 // CHECK2-NEXT: [[TMP94:%.*]] = load i32, i32* [[TMP93]], align 4
1561 // CHECK2-NEXT: [[TMP95:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
1562 // CHECK2-NEXT: [[TMP96:%.*]] = call i32 @__kmpc_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP94]], i32 2, i64 24, i8* [[TMP95]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
1563 // CHECK2-NEXT: switch i32 [[TMP96]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
1564 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
1565 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
1566 // CHECK2-NEXT: ]
1567 // CHECK2: .omp.reduction.case1:
1568 // CHECK2-NEXT: [[TMP97:%.*]] = load i32, i32* [[TMP0]], align 4
1569 // CHECK2-NEXT: [[TMP98:%.*]] = load i32, i32* [[ARGC1]], align 4
1570 // CHECK2-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP97]], [[TMP98]]
1571 // CHECK2-NEXT: store i32 [[ADD15]], i32* [[TMP0]], align 4
1572 // CHECK2-NEXT: [[TMP99:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP13]]
1573 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP99]]
1574 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE22:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1575 // CHECK2: omp.arraycpy.body:
1576 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1577 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST16:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT20:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1578 // CHECK2-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
1579 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP100]] to i32
1580 // CHECK2-NEXT: [[TMP101:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1581 // CHECK2-NEXT: [[CONV17:%.*]] = sext i8 [[TMP101]] to i32
1582 // CHECK2-NEXT: [[ADD18:%.*]] = add nsw i32 [[CONV]], [[CONV17]]
1583 // CHECK2-NEXT: [[CONV19:%.*]] = trunc i32 [[ADD18]] to i8
1584 // CHECK2-NEXT: store i8 [[CONV19]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], align 1
1585 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT20]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST16]], i32 1
1586 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1587 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE21:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT20]], [[TMP99]]
1588 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE21]], label [[OMP_ARRAYCPY_DONE22]], label [[OMP_ARRAYCPY_BODY]]
1589 // CHECK2: omp.arraycpy.done22:
1590 // CHECK2-NEXT: call void @__kmpc_end_reduce_nowait(%struct.ident_t* @[[GLOB4]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
1591 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1592 // CHECK2: .omp.reduction.case2:
1593 // CHECK2-NEXT: [[TMP102:%.*]] = load i32, i32* [[ARGC1]], align 4
1594 // CHECK2-NEXT: [[TMP103:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP102]] monotonic, align 4
1595 // CHECK2-NEXT: [[TMP104:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP13]]
1596 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY23:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP104]]
1597 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY23]], label [[OMP_ARRAYCPY_DONE36:%.*]], label [[OMP_ARRAYCPY_BODY24:%.*]]
1598 // CHECK2: omp.arraycpy.body24:
1599 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST25:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT34:%.*]], [[ATOMIC_EXIT:%.*]] ]
1600 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST26:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT33:%.*]], [[ATOMIC_EXIT]] ]
1601 // CHECK2-NEXT: [[TMP105:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
1602 // CHECK2-NEXT: [[CONV27:%.*]] = sext i8 [[TMP105]] to i32
1603 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]] monotonic, align 1
1604 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
1605 // CHECK2: atomic_cont:
1606 // CHECK2-NEXT: [[TMP106:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY24]] ], [ [[TMP111:%.*]], [[ATOMIC_CONT]] ]
1607 // CHECK2-NEXT: store i8 [[TMP106]], i8* [[_TMP28]], align 1
1608 // CHECK2-NEXT: [[TMP107:%.*]] = load i8, i8* [[_TMP28]], align 1
1609 // CHECK2-NEXT: [[CONV29:%.*]] = sext i8 [[TMP107]] to i32
1610 // CHECK2-NEXT: [[TMP108:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], align 1
1611 // CHECK2-NEXT: [[CONV30:%.*]] = sext i8 [[TMP108]] to i32
1612 // CHECK2-NEXT: [[ADD31:%.*]] = add nsw i32 [[CONV29]], [[CONV30]]
1613 // CHECK2-NEXT: [[CONV32:%.*]] = trunc i32 [[ADD31]] to i8
1614 // CHECK2-NEXT: store i8 [[CONV32]], i8* [[ATOMIC_TEMP]], align 1
1615 // CHECK2-NEXT: [[TMP109:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
1616 // CHECK2-NEXT: [[TMP110:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i8 [[TMP106]], i8 [[TMP109]] monotonic monotonic, align 1
1617 // CHECK2-NEXT: [[TMP111]] = extractvalue { i8, i1 } [[TMP110]], 0
1618 // CHECK2-NEXT: [[TMP112:%.*]] = extractvalue { i8, i1 } [[TMP110]], 1
1619 // CHECK2-NEXT: br i1 [[TMP112]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
1620 // CHECK2: atomic_exit:
1621 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT33]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST26]], i32 1
1622 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT34]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST25]], i32 1
1623 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE35:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT33]], [[TMP104]]
1624 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE35]], label [[OMP_ARRAYCPY_DONE36]], label [[OMP_ARRAYCPY_BODY24]]
1625 // CHECK2: omp.arraycpy.done36:
1626 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
1627 // CHECK2: .omp.reduction.default:
1628 // CHECK2-NEXT: [[TMP113:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1629 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP113]])
1630 // CHECK2-NEXT: ret void
1631 //
1632 //
1633 // CHECK2-LABEL: define {{[^@]+}}@.red_init..4
1634 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
1635 // CHECK2-NEXT: entry:
1636 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1637 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1638 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1639 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1640 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1641 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
1642 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
1643 // CHECK2-NEXT: ret void
1644 //
1645 //
1646 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..5
1647 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1648 // CHECK2-NEXT: entry:
1649 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1650 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1651 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1652 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1653 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1654 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
1655 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1656 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
1657 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
1658 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
1659 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
1660 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
1661 // CHECK2-NEXT: ret void
1662 //
1663 //
1664 // CHECK2-LABEL: define {{[^@]+}}@.red_init..6
1665 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
1666 // CHECK2-NEXT: entry:
1667 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1668 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1669 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1670 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1671 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1672 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
1673 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
1674 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
1675 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
1676 // CHECK2: omp.arrayinit.body:
1677 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
1678 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1679 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1680 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
1681 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
1682 // CHECK2: omp.arrayinit.done:
1683 // CHECK2-NEXT: ret void
1684 //
1685 //
1686 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..7
1687 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1688 // CHECK2-NEXT: entry:
1689 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1690 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1691 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1692 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1693 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
1694 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1695 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1696 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
1697 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
1698 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1699 // CHECK2: omp.arraycpy.body:
1700 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1701 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1702 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1703 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
1704 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1705 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
1706 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1707 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
1708 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1709 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1710 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1711 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
1712 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
1713 // CHECK2: omp.arraycpy.done4:
1714 // CHECK2-NEXT: ret void
1715 //
1716 //
1717 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
1718 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
1719 // CHECK2-NEXT: entry:
1720 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
1721 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
1722 // CHECK2-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
1723 // CHECK2-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
1724 // CHECK2-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
1725 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
1726 // CHECK2-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
1727 // CHECK2-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
1728 // CHECK2-NEXT: ret void
1729 //
1730 //
1731 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
1732 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
1733 // CHECK2-NEXT: entry:
1734 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
1735 // CHECK2-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
1736 // CHECK2-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
1737 // CHECK2-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
1738 // CHECK2-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
1739 // CHECK2-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
1740 // CHECK2-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
1741 // CHECK2-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
1742 // CHECK2-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
1743 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
1744 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
1745 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
1746 // CHECK2-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1747 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
1748 // CHECK2-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
1749 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
1750 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
1751 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
1752 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1753 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
1754 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
1755 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
1756 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
1757 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
1758 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
1759 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
1760 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
1761 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
1762 // CHECK2-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
1763 // CHECK2-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
1764 // 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
1765 // CHECK2-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
1766 // CHECK2-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
1767 // CHECK2-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
1768 // CHECK2-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
1769 // CHECK2-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
1770 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
1771 // CHECK2-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5]]
1772 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
1773 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
1774 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
1775 // CHECK2-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
1776 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
1777 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
1778 // CHECK2-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
1779 // CHECK2-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
1780 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1781 // CHECK2-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
1782 // CHECK2-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
1783 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
1784 // CHECK2-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
1785 // CHECK2-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
1786 // CHECK2-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
1787 // CHECK2-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
1788 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1789 // CHECK2-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
1790 // CHECK2-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
1791 // CHECK2-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
1792 // CHECK2-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
1793 // CHECK2-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
1794 // CHECK2-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
1795 // CHECK2-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
1796 // CHECK2-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1797 // CHECK2-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
1798 // CHECK2-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1799 // CHECK2-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
1800 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
1801 // CHECK2-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1802 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1803 // CHECK2-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1804 // CHECK2-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1805 // CHECK2-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1806 // CHECK2-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1807 // CHECK2-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1808 // CHECK2-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1809 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1810 // CHECK2-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1811 // CHECK2-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1812 // CHECK2-NEXT: ret i32 0
1813 //
1814 //
1815 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1816 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1817 // CHECK2-NEXT: entry:
1818 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1819 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1820 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1821 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1822 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1823 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1824 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1825 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1826 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1827 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1828 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1829 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1830 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1831 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1832 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1833 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1834 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1835 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1836 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1837 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1838 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1839 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1840 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1841 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1842 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1843 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1844 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1845 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1846 // CHECK2: omp.arraycpy.body:
1847 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1848 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1849 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1850 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1851 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1852 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1853 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1854 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1855 // CHECK2-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1856 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1857 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1858 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1859 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1860 // CHECK2: omp.arraycpy.done5:
1861 // CHECK2-NEXT: ret void
1862 //
1863 //
1864 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func.9
1865 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1866 // CHECK2-NEXT: entry:
1867 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1868 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1869 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1870 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1871 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1872 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1873 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1874 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1875 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1876 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1877 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1878 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1879 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1880 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1881 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1882 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1883 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1884 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1885 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1886 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1887 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1888 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1889 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1890 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1891 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1892 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1893 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1894 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1895 // CHECK2: omp.arraycpy.body:
1896 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1897 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1898 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1899 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1900 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1901 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1902 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1903 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1904 // CHECK2-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1905 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1906 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1907 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1908 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1909 // CHECK2: omp.arraycpy.done5:
1910 // CHECK2-NEXT: ret void
1911 //
1912