1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=50 -x c++ -triple x86_64-unknown-linux -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -fopenmp-version=50 -x c++ -triple x86_64-unknown-linux -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5
6 // RUN: %clang_cc1 -triple x86_64-unknown-linux -verify -fopenmp-simd -fopenmp-version=50 -x c++ -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
7 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
8 // RUN: %clang_cc1 -fopenmp-simd -fopenmp-version=50 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -debug-info-kind=limited -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // expected-no-diagnostics
10 #ifndef HEADER
11 #define HEADER
12
main(int argc,char ** argv)13 int main(int argc, char **argv) {
14 #pragma omp parallel
15 #pragma omp sections reduction(task, +: argc, argv[0:10][0:argc])
16 {
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: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
45 // CHECK1-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
46 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
47 // CHECK1-NEXT: ret i32 0
48 //
49 //
50 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
51 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
52 // CHECK1-NEXT: entry:
53 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
54 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
55 // CHECK1-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
56 // CHECK1-NEXT: [[ARGV_ADDR:%.*]] = alloca i8***, align 8
57 // CHECK1-NEXT: [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
58 // CHECK1-NEXT: [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
59 // CHECK1-NEXT: [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
60 // CHECK1-NEXT: [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
61 // CHECK1-NEXT: [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
62 // CHECK1-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
63 // CHECK1-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
64 // CHECK1-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
65 // CHECK1-NEXT: [[TMP:%.*]] = alloca i8**, align 8
66 // CHECK1-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
67 // CHECK1-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
68 // CHECK1-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
69 // CHECK1-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
70 // CHECK1-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
71 // CHECK1-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
72 // CHECK1-NEXT: [[_TMP24:%.*]] = alloca i8, align 1
73 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
74 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
75 // CHECK1-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
76 // CHECK1-NEXT: store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
77 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
78 // CHECK1-NEXT: [[TMP1:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
79 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
80 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
81 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
82 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
83 // CHECK1-NEXT: store i32 0, i32* [[ARGC1]], align 4
84 // CHECK1-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[TMP1]], align 8
85 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
86 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
87 // CHECK1-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP3]], i64 0
88 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP0]], align 4
89 // CHECK1-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
90 // CHECK1-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP5]]
91 // CHECK1-NEXT: [[TMP6:%.*]] = load i8**, i8*** [[TMP1]], align 8
92 // CHECK1-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP6]], i64 9
93 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
94 // CHECK1-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP7]], i64 [[LB_ADD_LEN]]
95 // CHECK1-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
96 // CHECK1-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
97 // CHECK1-NEXT: [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
98 // CHECK1-NEXT: [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
99 // CHECK1-NEXT: [[TMP12:%.*]] = add nuw i64 [[TMP11]], 1
100 // CHECK1-NEXT: [[TMP13:%.*]] = mul nuw i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
101 // CHECK1-NEXT: [[TMP14:%.*]] = call i8* @llvm.stacksave()
102 // CHECK1-NEXT: store i8* [[TMP14]], i8** [[SAVED_STACK]], align 8
103 // CHECK1-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP12]], align 16
104 // CHECK1-NEXT: store i64 [[TMP12]], i64* [[__VLA_EXPR0]], align 8
105 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP12]]
106 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP15]]
107 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
108 // CHECK1: omp.arrayinit.body:
109 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
110 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
111 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
112 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
113 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
114 // CHECK1: omp.arrayinit.done:
115 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[TMP1]], align 8
116 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
117 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
118 // CHECK1-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
119 // CHECK1-NEXT: [[TMP20:%.*]] = sub i64 [[TMP18]], [[TMP19]]
120 // CHECK1-NEXT: [[TMP21:%.*]] = sdiv exact i64 [[TMP20]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
121 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP21]]
122 // CHECK1-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
123 // CHECK1-NEXT: store i8* [[TMP22]], i8** [[_TMP5]], align 8
124 // 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
125 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
126 // CHECK1-NEXT: [[TMP24:%.*]] = bitcast i32* [[ARGC1]] to i8*
127 // CHECK1-NEXT: store i8* [[TMP24]], i8** [[TMP23]], align 8
128 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
129 // CHECK1-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP0]] to i8*
130 // CHECK1-NEXT: store i8* [[TMP26]], i8** [[TMP25]], align 8
131 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
132 // CHECK1-NEXT: store i64 4, i64* [[TMP27]], align 8
133 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
134 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP28]], align 8
135 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
136 // CHECK1-NEXT: store i8* null, i8** [[TMP29]], align 8
137 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
138 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP30]], align 8
139 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
140 // CHECK1-NEXT: [[TMP32:%.*]] = bitcast i32* [[TMP31]] to i8*
141 // CHECK1-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP32]], i8 0, i64 4, i1 false)
142 // 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
143 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
144 // CHECK1-NEXT: [[TMP34:%.*]] = load i8**, i8*** [[TMP1]], align 8
145 // CHECK1-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP34]], i64 0
146 // CHECK1-NEXT: [[TMP35:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
147 // CHECK1-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP35]], i64 0
148 // CHECK1-NEXT: [[TMP36:%.*]] = load i32, i32* [[TMP0]], align 4
149 // CHECK1-NEXT: [[TMP37:%.*]] = sext i32 [[TMP36]] to i64
150 // CHECK1-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP37]]
151 // CHECK1-NEXT: [[TMP38:%.*]] = load i8**, i8*** [[TMP1]], align 8
152 // CHECK1-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP38]], i64 9
153 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
154 // CHECK1-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP39]], i64 [[LB_ADD_LEN9]]
155 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP33]], align 8
156 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
157 // CHECK1-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP40]], align 8
158 // CHECK1-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
159 // CHECK1-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
160 // CHECK1-NEXT: [[TMP43:%.*]] = sub i64 [[TMP41]], [[TMP42]]
161 // CHECK1-NEXT: [[TMP44:%.*]] = sdiv exact i64 [[TMP43]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
162 // CHECK1-NEXT: [[TMP45:%.*]] = add nuw i64 [[TMP44]], 1
163 // CHECK1-NEXT: [[TMP46:%.*]] = mul nuw i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
164 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
165 // CHECK1-NEXT: store i64 [[TMP46]], i64* [[TMP47]], align 8
166 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
167 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP48]], align 8
168 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
169 // CHECK1-NEXT: store i8* null, i8** [[TMP49]], align 8
170 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
171 // CHECK1-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP50]], align 8
172 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
173 // CHECK1-NEXT: store i32 1, i32* [[TMP51]], align 8
174 // CHECK1-NEXT: [[TMP52:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
175 // CHECK1-NEXT: [[TMP53:%.*]] = load i32, i32* [[TMP52]], align 4
176 // CHECK1-NEXT: [[TMP54:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
177 // CHECK1-NEXT: [[TMP55:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], i32 1, i32 2, i8* [[TMP54]])
178 // CHECK1-NEXT: store i8* [[TMP55]], i8** [[DOTTASK_RED_]], align 8
179 // CHECK1-NEXT: [[TMP56:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
180 // CHECK1-NEXT: [[TMP57:%.*]] = load i32, i32* [[TMP56]], align 4
181 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP57]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
182 // CHECK1-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
183 // CHECK1-NEXT: [[TMP59:%.*]] = icmp slt i32 [[TMP58]], 0
184 // CHECK1-NEXT: [[TMP60:%.*]] = select i1 [[TMP59]], i32 [[TMP58]], i32 0
185 // CHECK1-NEXT: store i32 [[TMP60]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
186 // CHECK1-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
187 // CHECK1-NEXT: store i32 [[TMP61]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
188 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
189 // CHECK1: omp.inner.for.cond:
190 // CHECK1-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
191 // CHECK1-NEXT: [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
192 // CHECK1-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP62]], [[TMP63]]
193 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
194 // CHECK1: omp.inner.for.body:
195 // CHECK1-NEXT: [[TMP64:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
196 // CHECK1-NEXT: switch i32 [[TMP64]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
197 // CHECK1-NEXT: i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
198 // CHECK1-NEXT: ]
199 // CHECK1: .omp.sections.case:
200 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
201 // CHECK1-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP65]], align 8
202 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
203 // CHECK1-NEXT: store i32* [[ARGC1]], i32** [[TMP66]], align 8
204 // CHECK1-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
205 // CHECK1-NEXT: [[TMP68:%.*]] = load i8**, i8*** [[TMP]], align 8
206 // CHECK1-NEXT: store i8** [[TMP68]], i8*** [[TMP67]], align 8
207 // CHECK1-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
208 // CHECK1-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
209 // 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*)*))
210 // CHECK1-NEXT: [[TMP72:%.*]] = bitcast i8* [[TMP71]] to %struct.kmp_task_t_with_privates*
211 // CHECK1-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 0
212 // CHECK1-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP73]], i32 0, i32 0
213 // CHECK1-NEXT: [[TMP75:%.*]] = load i8*, i8** [[TMP74]], align 8
214 // CHECK1-NEXT: [[TMP76:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
215 // CHECK1-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP75]], i8* align 8 [[TMP76]], i64 24, i1 false)
216 // CHECK1-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 1
217 // CHECK1-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP77]], i32 0, i32 0
218 // CHECK1-NEXT: [[TMP79:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
219 // CHECK1-NEXT: store i8* [[TMP79]], i8** [[TMP78]], align 8
220 // CHECK1-NEXT: [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
221 // CHECK1-NEXT: [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
222 // CHECK1-NEXT: [[TMP82:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i8* [[TMP71]])
223 // CHECK1-NEXT: br label [[DOTOMP_SECTIONS_EXIT]]
224 // CHECK1: .omp.sections.exit:
225 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
226 // CHECK1: omp.inner.for.inc:
227 // CHECK1-NEXT: [[TMP83:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
228 // CHECK1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP83]], 1
229 // CHECK1-NEXT: store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
230 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
231 // CHECK1: omp.inner.for.end:
232 // CHECK1-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
233 // CHECK1-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
234 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]])
235 // CHECK1-NEXT: [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
236 // CHECK1-NEXT: [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
237 // CHECK1-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP87]], i32 1)
238 // CHECK1-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
239 // CHECK1-NEXT: [[TMP89:%.*]] = bitcast i32* [[ARGC1]] to i8*
240 // CHECK1-NEXT: store i8* [[TMP89]], i8** [[TMP88]], align 8
241 // CHECK1-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
242 // CHECK1-NEXT: store i8* [[VLA]], i8** [[TMP90]], align 8
243 // CHECK1-NEXT: [[TMP91:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
244 // CHECK1-NEXT: [[TMP92:%.*]] = inttoptr i64 [[TMP12]] to i8*
245 // CHECK1-NEXT: store i8* [[TMP92]], i8** [[TMP91]], align 8
246 // CHECK1-NEXT: [[TMP93:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
247 // CHECK1-NEXT: [[TMP94:%.*]] = load i32, i32* [[TMP93]], align 4
248 // CHECK1-NEXT: [[TMP95:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
249 // CHECK1-NEXT: [[TMP96:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP94]], i32 2, i64 24, i8* [[TMP95]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
250 // CHECK1-NEXT: switch i32 [[TMP96]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
251 // CHECK1-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
252 // CHECK1-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
253 // CHECK1-NEXT: ]
254 // CHECK1: .omp.reduction.case1:
255 // CHECK1-NEXT: [[TMP97:%.*]] = load i32, i32* [[TMP0]], align 4
256 // CHECK1-NEXT: [[TMP98:%.*]] = load i32, i32* [[ARGC1]], align 4
257 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP97]], [[TMP98]]
258 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
259 // CHECK1-NEXT: [[TMP99:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
260 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP99]]
261 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
262 // CHECK1: omp.arraycpy.body:
263 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
264 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
265 // CHECK1-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
266 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP100]] to i32
267 // CHECK1-NEXT: [[TMP101:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
268 // CHECK1-NEXT: [[CONV13:%.*]] = sext i8 [[TMP101]] to i32
269 // CHECK1-NEXT: [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
270 // CHECK1-NEXT: [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
271 // CHECK1-NEXT: store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
272 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
273 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
274 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP99]]
275 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
276 // CHECK1: omp.arraycpy.done18:
277 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
278 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
279 // CHECK1: .omp.reduction.case2:
280 // CHECK1-NEXT: [[TMP102:%.*]] = load i32, i32* [[ARGC1]], align 4
281 // CHECK1-NEXT: [[TMP103:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP102]] monotonic, align 4
282 // CHECK1-NEXT: [[TMP104:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
283 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP104]]
284 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
285 // CHECK1: omp.arraycpy.body20:
286 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
287 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
288 // CHECK1-NEXT: [[TMP105:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
289 // CHECK1-NEXT: [[CONV23:%.*]] = sext i8 [[TMP105]] to i32
290 // CHECK1-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
291 // CHECK1-NEXT: br label [[ATOMIC_CONT:%.*]]
292 // CHECK1: atomic_cont:
293 // CHECK1-NEXT: [[TMP106:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP111:%.*]], [[ATOMIC_CONT]] ]
294 // CHECK1-NEXT: store i8 [[TMP106]], i8* [[_TMP24]], align 1
295 // CHECK1-NEXT: [[TMP107:%.*]] = load i8, i8* [[_TMP24]], align 1
296 // CHECK1-NEXT: [[CONV25:%.*]] = sext i8 [[TMP107]] to i32
297 // CHECK1-NEXT: [[TMP108:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
298 // CHECK1-NEXT: [[CONV26:%.*]] = sext i8 [[TMP108]] to i32
299 // CHECK1-NEXT: [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
300 // CHECK1-NEXT: [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
301 // CHECK1-NEXT: store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
302 // CHECK1-NEXT: [[TMP109:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
303 // CHECK1-NEXT: [[TMP110:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP106]], i8 [[TMP109]] monotonic monotonic, align 1
304 // CHECK1-NEXT: [[TMP111]] = extractvalue { i8, i1 } [[TMP110]], 0
305 // CHECK1-NEXT: [[TMP112:%.*]] = extractvalue { i8, i1 } [[TMP110]], 1
306 // CHECK1-NEXT: br i1 [[TMP112]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
307 // CHECK1: atomic_exit:
308 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
309 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
310 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP104]]
311 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
312 // CHECK1: omp.arraycpy.done32:
313 // CHECK1-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
314 // CHECK1-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
315 // CHECK1: .omp.reduction.default:
316 // CHECK1-NEXT: [[TMP113:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
317 // CHECK1-NEXT: call void @llvm.stackrestore(i8* [[TMP113]])
318 // CHECK1-NEXT: [[TMP114:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
319 // CHECK1-NEXT: [[TMP115:%.*]] = load i32, i32* [[TMP114]], align 4
320 // CHECK1-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP115]])
321 // CHECK1-NEXT: ret void
322 //
323 //
324 // CHECK1-LABEL: define {{[^@]+}}@.red_init.
325 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
326 // CHECK1-NEXT: entry:
327 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
328 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
329 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
330 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
331 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
332 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
333 // CHECK1-NEXT: store i32 0, i32* [[TMP3]], align 8
334 // CHECK1-NEXT: ret void
335 //
336 //
337 // CHECK1-LABEL: define {{[^@]+}}@.red_comb.
338 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
339 // CHECK1-NEXT: entry:
340 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
341 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
342 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
343 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
344 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
345 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
346 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
347 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
348 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
349 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
350 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
351 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
352 // CHECK1-NEXT: ret void
353 //
354 //
355 // CHECK1-LABEL: define {{[^@]+}}@.red_init..1
356 // CHECK1-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
357 // CHECK1-NEXT: entry:
358 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
359 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
360 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
361 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
362 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
363 // CHECK1-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
364 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
365 // CHECK1-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
366 // CHECK1-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
367 // CHECK1: omp.arrayinit.body:
368 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
369 // CHECK1-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
370 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
371 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
372 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
373 // CHECK1: omp.arrayinit.done:
374 // CHECK1-NEXT: ret void
375 //
376 //
377 // CHECK1-LABEL: define {{[^@]+}}@.red_comb..2
378 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
379 // CHECK1-NEXT: entry:
380 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
381 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
382 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
383 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
384 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
385 // CHECK1-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
386 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
387 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
388 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
389 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
390 // CHECK1: omp.arraycpy.body:
391 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
392 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
393 // CHECK1-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
394 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
395 // CHECK1-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
396 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
397 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
398 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
399 // CHECK1-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
400 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
401 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
402 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
403 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
404 // CHECK1: omp.arraycpy.done4:
405 // CHECK1-NEXT: ret void
406 //
407 //
408 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_privates_map.
409 // CHECK1-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
410 // CHECK1-NEXT: entry:
411 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
412 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
413 // CHECK1-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
414 // CHECK1-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
415 // CHECK1-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
416 // CHECK1-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
417 // CHECK1-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
418 // CHECK1-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
419 // CHECK1-NEXT: ret void
420 //
421 //
422 // CHECK1-LABEL: define {{[^@]+}}@.omp_task_entry.
423 // CHECK1-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
424 // CHECK1-NEXT: entry:
425 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
426 // CHECK1-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
427 // CHECK1-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
428 // CHECK1-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
429 // CHECK1-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
430 // CHECK1-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
431 // CHECK1-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
432 // CHECK1-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
433 // CHECK1-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
434 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
435 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
436 // CHECK1-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
437 // CHECK1-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
438 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
439 // CHECK1-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
440 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
441 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
442 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
443 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
444 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
445 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
446 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
447 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
448 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
449 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
450 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
451 // CHECK1-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
452 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
453 // CHECK1-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
454 // CHECK1-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
455 // 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
456 // CHECK1-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
457 // CHECK1-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
458 // CHECK1-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
459 // CHECK1-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
460 // CHECK1-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
461 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
462 // CHECK1-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
463 // CHECK1-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
464 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
465 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
466 // CHECK1-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
467 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
468 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
469 // CHECK1-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
470 // CHECK1-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
471 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
472 // CHECK1-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
473 // CHECK1-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
474 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
475 // CHECK1-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
476 // CHECK1-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
477 // CHECK1-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
478 // CHECK1-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
479 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
480 // CHECK1-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
481 // CHECK1-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
482 // CHECK1-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
483 // CHECK1-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
484 // CHECK1-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
485 // CHECK1-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
486 // CHECK1-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
487 // CHECK1-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
488 // CHECK1-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
489 // CHECK1-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
490 // CHECK1-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
491 // CHECK1-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
492 // CHECK1-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
493 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
494 // CHECK1-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
495 // CHECK1-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
496 // CHECK1-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
497 // CHECK1-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
498 // CHECK1-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
499 // CHECK1-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
500 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
501 // CHECK1-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
502 // CHECK1-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
503 // CHECK1-NEXT: ret i32 0
504 //
505 //
506 // CHECK1-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
507 // CHECK1-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
508 // CHECK1-NEXT: entry:
509 // CHECK1-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
510 // CHECK1-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
511 // CHECK1-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
512 // CHECK1-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
513 // CHECK1-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
514 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
515 // CHECK1-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
516 // CHECK1-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
517 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
518 // CHECK1-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
519 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
520 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
521 // CHECK1-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
522 // CHECK1-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
523 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
524 // CHECK1-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
525 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
526 // CHECK1-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
527 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
528 // CHECK1-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
529 // CHECK1-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
530 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
531 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
532 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
533 // CHECK1-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
534 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
535 // CHECK1-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
536 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
537 // CHECK1: omp.arraycpy.body:
538 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
539 // CHECK1-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
540 // CHECK1-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
541 // CHECK1-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
542 // CHECK1-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
543 // CHECK1-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
544 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
545 // CHECK1-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
546 // CHECK1-NEXT: store i8 [[CONV4]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
547 // CHECK1-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
548 // CHECK1-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
549 // CHECK1-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
550 // CHECK1-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
551 // CHECK1: omp.arraycpy.done5:
552 // CHECK1-NEXT: ret void
553 //
554 //
555 // CHECK2-LABEL: define {{[^@]+}}@main
556 // CHECK2-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
557 // CHECK2-NEXT: entry:
558 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
559 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
560 // CHECK2-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
561 // CHECK2-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
562 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i8***)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[ARGC_ADDR]], i8*** [[ARGV_ADDR]])
563 // CHECK2-NEXT: ret i32 0
564 //
565 //
566 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
567 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[ARGC:%.*]], i8*** nonnull align 8 dereferenceable(8) [[ARGV:%.*]]) #[[ATTR1:[0-9]+]] {
568 // CHECK2-NEXT: entry:
569 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
570 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
571 // CHECK2-NEXT: [[ARGC_ADDR:%.*]] = alloca i32*, align 8
572 // CHECK2-NEXT: [[ARGV_ADDR:%.*]] = alloca i8***, align 8
573 // CHECK2-NEXT: [[DOTOMP_SECTIONS_LB_:%.*]] = alloca i32, align 4
574 // CHECK2-NEXT: [[DOTOMP_SECTIONS_UB_:%.*]] = alloca i32, align 4
575 // CHECK2-NEXT: [[DOTOMP_SECTIONS_ST_:%.*]] = alloca i32, align 4
576 // CHECK2-NEXT: [[DOTOMP_SECTIONS_IL_:%.*]] = alloca i32, align 4
577 // CHECK2-NEXT: [[DOTOMP_SECTIONS_IV_:%.*]] = alloca i32, align 4
578 // CHECK2-NEXT: [[ARGC1:%.*]] = alloca i32, align 4
579 // CHECK2-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
580 // CHECK2-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
581 // CHECK2-NEXT: [[TMP:%.*]] = alloca i8**, align 8
582 // CHECK2-NEXT: [[_TMP5:%.*]] = alloca i8*, align 8
583 // CHECK2-NEXT: [[DOTRD_INPUT_:%.*]] = alloca [2 x %struct.kmp_taskred_input_t], align 8
584 // CHECK2-NEXT: [[DOTTASK_RED_:%.*]] = alloca i8*, align 8
585 // CHECK2-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
586 // CHECK2-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [3 x i8*], align 8
587 // CHECK2-NEXT: [[ATOMIC_TEMP:%.*]] = alloca i8, align 1
588 // CHECK2-NEXT: [[_TMP24:%.*]] = alloca i8, align 1
589 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
590 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
591 // CHECK2-NEXT: store i32* [[ARGC]], i32** [[ARGC_ADDR]], align 8
592 // CHECK2-NEXT: store i8*** [[ARGV]], i8**** [[ARGV_ADDR]], align 8
593 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[ARGC_ADDR]], align 8
594 // CHECK2-NEXT: [[TMP1:%.*]] = load i8***, i8**** [[ARGV_ADDR]], align 8
595 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_LB_]], align 4
596 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_UB_]], align 4
597 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_SECTIONS_ST_]], align 4
598 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_SECTIONS_IL_]], align 4
599 // CHECK2-NEXT: store i32 0, i32* [[ARGC1]], align 4
600 // CHECK2-NEXT: [[TMP2:%.*]] = load i8**, i8*** [[TMP1]], align 8
601 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8*, i8** [[TMP2]], i64 0
602 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[ARRAYIDX]], align 8
603 // CHECK2-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, i8* [[TMP3]], i64 0
604 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP0]], align 4
605 // CHECK2-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
606 // CHECK2-NEXT: [[LB_ADD_LEN:%.*]] = add nsw i64 -1, [[TMP5]]
607 // CHECK2-NEXT: [[TMP6:%.*]] = load i8**, i8*** [[TMP1]], align 8
608 // CHECK2-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i8*, i8** [[TMP6]], i64 9
609 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[ARRAYIDX3]], align 8
610 // CHECK2-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[TMP7]], i64 [[LB_ADD_LEN]]
611 // CHECK2-NEXT: [[TMP8:%.*]] = ptrtoint i8* [[ARRAYIDX4]] to i64
612 // CHECK2-NEXT: [[TMP9:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
613 // CHECK2-NEXT: [[TMP10:%.*]] = sub i64 [[TMP8]], [[TMP9]]
614 // CHECK2-NEXT: [[TMP11:%.*]] = sdiv exact i64 [[TMP10]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
615 // CHECK2-NEXT: [[TMP12:%.*]] = add nuw i64 [[TMP11]], 1
616 // CHECK2-NEXT: [[TMP13:%.*]] = mul nuw i64 [[TMP12]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
617 // CHECK2-NEXT: [[TMP14:%.*]] = call i8* @llvm.stacksave()
618 // CHECK2-NEXT: store i8* [[TMP14]], i8** [[SAVED_STACK]], align 8
619 // CHECK2-NEXT: [[VLA:%.*]] = alloca i8, i64 [[TMP12]], align 16
620 // CHECK2-NEXT: store i64 [[TMP12]], i64* [[__VLA_EXPR0]], align 8
621 // CHECK2-NEXT: [[TMP15:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP12]]
622 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[VLA]], [[TMP15]]
623 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
624 // CHECK2: omp.arrayinit.body:
625 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
626 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
627 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
628 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP15]]
629 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
630 // CHECK2: omp.arrayinit.done:
631 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[TMP1]], align 8
632 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
633 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
634 // CHECK2-NEXT: [[TMP19:%.*]] = ptrtoint i8* [[ARRAYIDX2]] to i64
635 // CHECK2-NEXT: [[TMP20:%.*]] = sub i64 [[TMP18]], [[TMP19]]
636 // CHECK2-NEXT: [[TMP21:%.*]] = sdiv exact i64 [[TMP20]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
637 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr i8, i8* [[VLA]], i64 [[TMP21]]
638 // CHECK2-NEXT: store i8** [[_TMP5]], i8*** [[TMP]], align 8
639 // CHECK2-NEXT: store i8* [[TMP22]], i8** [[_TMP5]], align 8
640 // 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
641 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T:%.*]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 0
642 // CHECK2-NEXT: [[TMP24:%.*]] = bitcast i32* [[ARGC1]] to i8*
643 // CHECK2-NEXT: store i8* [[TMP24]], i8** [[TMP23]], align 8
644 // CHECK2-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
645 // CHECK2-NEXT: [[TMP26:%.*]] = bitcast i32* [[TMP0]] to i8*
646 // CHECK2-NEXT: store i8* [[TMP26]], i8** [[TMP25]], align 8
647 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
648 // CHECK2-NEXT: store i64 4, i64* [[TMP27]], align 8
649 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
650 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init. to i8*), i8** [[TMP28]], align 8
651 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
652 // CHECK2-NEXT: store i8* null, i8** [[TMP29]], align 8
653 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
654 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb. to i8*), i8** [[TMP30]], align 8
655 // CHECK2-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 6
656 // CHECK2-NEXT: [[TMP32:%.*]] = bitcast i32* [[TMP31]] to i8*
657 // CHECK2-NEXT: call void @llvm.memset.p0i8.i64(i8* align 8 [[TMP32]], i8 0, i64 4, i1 false)
658 // 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
659 // CHECK2-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 0
660 // CHECK2-NEXT: [[TMP34:%.*]] = load i8**, i8*** [[TMP1]], align 8
661 // CHECK2-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i8*, i8** [[TMP34]], i64 0
662 // CHECK2-NEXT: [[TMP35:%.*]] = load i8*, i8** [[ARRAYIDX7]], align 8
663 // CHECK2-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[TMP35]], i64 0
664 // CHECK2-NEXT: [[TMP36:%.*]] = load i32, i32* [[TMP0]], align 4
665 // CHECK2-NEXT: [[TMP37:%.*]] = sext i32 [[TMP36]] to i64
666 // CHECK2-NEXT: [[LB_ADD_LEN9:%.*]] = add nsw i64 -1, [[TMP37]]
667 // CHECK2-NEXT: [[TMP38:%.*]] = load i8**, i8*** [[TMP1]], align 8
668 // CHECK2-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i8*, i8** [[TMP38]], i64 9
669 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[ARRAYIDX10]], align 8
670 // CHECK2-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i8, i8* [[TMP39]], i64 [[LB_ADD_LEN9]]
671 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP33]], align 8
672 // CHECK2-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 1
673 // CHECK2-NEXT: store i8* [[ARRAYIDX8]], i8** [[TMP40]], align 8
674 // CHECK2-NEXT: [[TMP41:%.*]] = ptrtoint i8* [[ARRAYIDX11]] to i64
675 // CHECK2-NEXT: [[TMP42:%.*]] = ptrtoint i8* [[ARRAYIDX8]] to i64
676 // CHECK2-NEXT: [[TMP43:%.*]] = sub i64 [[TMP41]], [[TMP42]]
677 // CHECK2-NEXT: [[TMP44:%.*]] = sdiv exact i64 [[TMP43]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
678 // CHECK2-NEXT: [[TMP45:%.*]] = add nuw i64 [[TMP44]], 1
679 // CHECK2-NEXT: [[TMP46:%.*]] = mul nuw i64 [[TMP45]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
680 // CHECK2-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 2
681 // CHECK2-NEXT: store i64 [[TMP46]], i64* [[TMP47]], align 8
682 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 3
683 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_init..1 to i8*), i8** [[TMP48]], align 8
684 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 4
685 // CHECK2-NEXT: store i8* null, i8** [[TMP49]], align 8
686 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 5
687 // CHECK2-NEXT: store i8* bitcast (void (i8*, i8*)* @.red_comb..2 to i8*), i8** [[TMP50]], align 8
688 // CHECK2-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASKRED_INPUT_T]], %struct.kmp_taskred_input_t* [[DOTRD_INPUT_GEP_6]], i32 0, i32 6
689 // CHECK2-NEXT: store i32 1, i32* [[TMP51]], align 8
690 // CHECK2-NEXT: [[TMP52:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
691 // CHECK2-NEXT: [[TMP53:%.*]] = load i32, i32* [[TMP52]], align 4
692 // CHECK2-NEXT: [[TMP54:%.*]] = bitcast [2 x %struct.kmp_taskred_input_t]* [[DOTRD_INPUT_]] to i8*
693 // CHECK2-NEXT: [[TMP55:%.*]] = call i8* @__kmpc_taskred_modifier_init(%struct.ident_t* @[[GLOB1]], i32 [[TMP53]], i32 1, i32 2, i8* [[TMP54]])
694 // CHECK2-NEXT: store i8* [[TMP55]], i8** [[DOTTASK_RED_]], align 8
695 // CHECK2-NEXT: [[TMP56:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
696 // CHECK2-NEXT: [[TMP57:%.*]] = load i32, i32* [[TMP56]], align 4
697 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP57]], i32 34, i32* [[DOTOMP_SECTIONS_IL_]], i32* [[DOTOMP_SECTIONS_LB_]], i32* [[DOTOMP_SECTIONS_UB_]], i32* [[DOTOMP_SECTIONS_ST_]], i32 1, i32 1)
698 // CHECK2-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
699 // CHECK2-NEXT: [[TMP59:%.*]] = icmp slt i32 [[TMP58]], 0
700 // CHECK2-NEXT: [[TMP60:%.*]] = select i1 [[TMP59]], i32 [[TMP58]], i32 0
701 // CHECK2-NEXT: store i32 [[TMP60]], i32* [[DOTOMP_SECTIONS_UB_]], align 4
702 // CHECK2-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_LB_]], align 4
703 // CHECK2-NEXT: store i32 [[TMP61]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
704 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
705 // CHECK2: omp.inner.for.cond:
706 // CHECK2-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
707 // CHECK2-NEXT: [[TMP63:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_UB_]], align 4
708 // CHECK2-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP62]], [[TMP63]]
709 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
710 // CHECK2: omp.inner.for.body:
711 // CHECK2-NEXT: [[TMP64:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
712 // CHECK2-NEXT: switch i32 [[TMP64]], label [[DOTOMP_SECTIONS_EXIT:%.*]] [
713 // CHECK2-NEXT: i32 0, label [[DOTOMP_SECTIONS_CASE:%.*]]
714 // CHECK2-NEXT: ]
715 // CHECK2: .omp.sections.case:
716 // CHECK2-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 0
717 // CHECK2-NEXT: store i8** [[DOTTASK_RED_]], i8*** [[TMP65]], align 8
718 // CHECK2-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 1
719 // CHECK2-NEXT: store i32* [[ARGC1]], i32** [[TMP66]], align 8
720 // CHECK2-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[AGG_CAPTURED]], i32 0, i32 2
721 // CHECK2-NEXT: [[TMP68:%.*]] = load i8**, i8*** [[TMP]], align 8
722 // CHECK2-NEXT: store i8** [[TMP68]], i8*** [[TMP67]], align 8
723 // CHECK2-NEXT: [[TMP69:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
724 // CHECK2-NEXT: [[TMP70:%.*]] = load i32, i32* [[TMP69]], align 4
725 // 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*)*))
726 // CHECK2-NEXT: [[TMP72:%.*]] = bitcast i8* [[TMP71]] to %struct.kmp_task_t_with_privates*
727 // CHECK2-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 0
728 // CHECK2-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP73]], i32 0, i32 0
729 // CHECK2-NEXT: [[TMP75:%.*]] = load i8*, i8** [[TMP74]], align 8
730 // CHECK2-NEXT: [[TMP76:%.*]] = bitcast %struct.anon* [[AGG_CAPTURED]] to i8*
731 // CHECK2-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP75]], i8* align 8 [[TMP76]], i64 24, i1 false)
732 // CHECK2-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP72]], i32 0, i32 1
733 // CHECK2-NEXT: [[TMP78:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP77]], i32 0, i32 0
734 // CHECK2-NEXT: [[TMP79:%.*]] = load i8*, i8** [[DOTTASK_RED_]], align 8
735 // CHECK2-NEXT: store i8* [[TMP79]], i8** [[TMP78]], align 8
736 // CHECK2-NEXT: [[TMP80:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
737 // CHECK2-NEXT: [[TMP81:%.*]] = load i32, i32* [[TMP80]], align 4
738 // CHECK2-NEXT: [[TMP82:%.*]] = call i32 @__kmpc_omp_task(%struct.ident_t* @[[GLOB1]], i32 [[TMP81]], i8* [[TMP71]])
739 // CHECK2-NEXT: br label [[DOTOMP_SECTIONS_EXIT]]
740 // CHECK2: .omp.sections.exit:
741 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
742 // CHECK2: omp.inner.for.inc:
743 // CHECK2-NEXT: [[TMP83:%.*]] = load i32, i32* [[DOTOMP_SECTIONS_IV_]], align 4
744 // CHECK2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP83]], 1
745 // CHECK2-NEXT: store i32 [[INC]], i32* [[DOTOMP_SECTIONS_IV_]], align 4
746 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
747 // CHECK2: omp.inner.for.end:
748 // CHECK2-NEXT: [[TMP84:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
749 // CHECK2-NEXT: [[TMP85:%.*]] = load i32, i32* [[TMP84]], align 4
750 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB2]], i32 [[TMP85]])
751 // CHECK2-NEXT: [[TMP86:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
752 // CHECK2-NEXT: [[TMP87:%.*]] = load i32, i32* [[TMP86]], align 4
753 // CHECK2-NEXT: call void @__kmpc_task_reduction_modifier_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP87]], i32 1)
754 // CHECK2-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 0
755 // CHECK2-NEXT: [[TMP89:%.*]] = bitcast i32* [[ARGC1]] to i8*
756 // CHECK2-NEXT: store i8* [[TMP89]], i8** [[TMP88]], align 8
757 // CHECK2-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 1
758 // CHECK2-NEXT: store i8* [[VLA]], i8** [[TMP90]], align 8
759 // CHECK2-NEXT: [[TMP91:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]], i64 0, i64 2
760 // CHECK2-NEXT: [[TMP92:%.*]] = inttoptr i64 [[TMP12]] to i8*
761 // CHECK2-NEXT: store i8* [[TMP92]], i8** [[TMP91]], align 8
762 // CHECK2-NEXT: [[TMP93:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
763 // CHECK2-NEXT: [[TMP94:%.*]] = load i32, i32* [[TMP93]], align 4
764 // CHECK2-NEXT: [[TMP95:%.*]] = bitcast [3 x i8*]* [[DOTOMP_REDUCTION_RED_LIST]] to i8*
765 // CHECK2-NEXT: [[TMP96:%.*]] = call i32 @__kmpc_reduce(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP94]], i32 2, i64 24, i8* [[TMP95]], void (i8*, i8*)* @.omp.reduction.reduction_func, [8 x i32]* @.gomp_critical_user_.reduction.var)
766 // CHECK2-NEXT: switch i32 [[TMP96]], label [[DOTOMP_REDUCTION_DEFAULT:%.*]] [
767 // CHECK2-NEXT: i32 1, label [[DOTOMP_REDUCTION_CASE1:%.*]]
768 // CHECK2-NEXT: i32 2, label [[DOTOMP_REDUCTION_CASE2:%.*]]
769 // CHECK2-NEXT: ]
770 // CHECK2: .omp.reduction.case1:
771 // CHECK2-NEXT: [[TMP97:%.*]] = load i32, i32* [[TMP0]], align 4
772 // CHECK2-NEXT: [[TMP98:%.*]] = load i32, i32* [[ARGC1]], align 4
773 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP97]], [[TMP98]]
774 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP0]], align 4
775 // CHECK2-NEXT: [[TMP99:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
776 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP99]]
777 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE18:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
778 // CHECK2: omp.arraycpy.body:
779 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
780 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST12:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE1]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT16:%.*]], [[OMP_ARRAYCPY_BODY]] ]
781 // CHECK2-NEXT: [[TMP100:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
782 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP100]] to i32
783 // CHECK2-NEXT: [[TMP101:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
784 // CHECK2-NEXT: [[CONV13:%.*]] = sext i8 [[TMP101]] to i32
785 // CHECK2-NEXT: [[ADD14:%.*]] = add nsw i32 [[CONV]], [[CONV13]]
786 // CHECK2-NEXT: [[CONV15:%.*]] = trunc i32 [[ADD14]] to i8
787 // CHECK2-NEXT: store i8 [[CONV15]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], align 1
788 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT16]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST12]], i32 1
789 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
790 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE17:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT16]], [[TMP99]]
791 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE17]], label [[OMP_ARRAYCPY_DONE18]], label [[OMP_ARRAYCPY_BODY]]
792 // CHECK2: omp.arraycpy.done18:
793 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
794 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
795 // CHECK2: .omp.reduction.case2:
796 // CHECK2-NEXT: [[TMP102:%.*]] = load i32, i32* [[ARGC1]], align 4
797 // CHECK2-NEXT: [[TMP103:%.*]] = atomicrmw add i32* [[TMP0]], i32 [[TMP102]] monotonic, align 4
798 // CHECK2-NEXT: [[TMP104:%.*]] = getelementptr i8, i8* [[ARRAYIDX2]], i64 [[TMP12]]
799 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY19:%.*]] = icmp eq i8* [[ARRAYIDX2]], [[TMP104]]
800 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY19]], label [[OMP_ARRAYCPY_DONE32:%.*]], label [[OMP_ARRAYCPY_BODY20:%.*]]
801 // CHECK2: omp.arraycpy.body20:
802 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST21:%.*]] = phi i8* [ [[VLA]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT30:%.*]], [[ATOMIC_EXIT:%.*]] ]
803 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST22:%.*]] = phi i8* [ [[ARRAYIDX2]], [[DOTOMP_REDUCTION_CASE2]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT29:%.*]], [[ATOMIC_EXIT]] ]
804 // CHECK2-NEXT: [[TMP105:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
805 // CHECK2-NEXT: [[CONV23:%.*]] = sext i8 [[TMP105]] to i32
806 // CHECK2-NEXT: [[ATOMIC_LOAD:%.*]] = load atomic i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]] monotonic, align 1
807 // CHECK2-NEXT: br label [[ATOMIC_CONT:%.*]]
808 // CHECK2: atomic_cont:
809 // CHECK2-NEXT: [[TMP106:%.*]] = phi i8 [ [[ATOMIC_LOAD]], [[OMP_ARRAYCPY_BODY20]] ], [ [[TMP111:%.*]], [[ATOMIC_CONT]] ]
810 // CHECK2-NEXT: store i8 [[TMP106]], i8* [[_TMP24]], align 1
811 // CHECK2-NEXT: [[TMP107:%.*]] = load i8, i8* [[_TMP24]], align 1
812 // CHECK2-NEXT: [[CONV25:%.*]] = sext i8 [[TMP107]] to i32
813 // CHECK2-NEXT: [[TMP108:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], align 1
814 // CHECK2-NEXT: [[CONV26:%.*]] = sext i8 [[TMP108]] to i32
815 // CHECK2-NEXT: [[ADD27:%.*]] = add nsw i32 [[CONV25]], [[CONV26]]
816 // CHECK2-NEXT: [[CONV28:%.*]] = trunc i32 [[ADD27]] to i8
817 // CHECK2-NEXT: store i8 [[CONV28]], i8* [[ATOMIC_TEMP]], align 1
818 // CHECK2-NEXT: [[TMP109:%.*]] = load i8, i8* [[ATOMIC_TEMP]], align 1
819 // CHECK2-NEXT: [[TMP110:%.*]] = cmpxchg i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i8 [[TMP106]], i8 [[TMP109]] monotonic monotonic, align 1
820 // CHECK2-NEXT: [[TMP111]] = extractvalue { i8, i1 } [[TMP110]], 0
821 // CHECK2-NEXT: [[TMP112:%.*]] = extractvalue { i8, i1 } [[TMP110]], 1
822 // CHECK2-NEXT: br i1 [[TMP112]], label [[ATOMIC_EXIT]], label [[ATOMIC_CONT]]
823 // CHECK2: atomic_exit:
824 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT29]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST22]], i32 1
825 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT30]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST21]], i32 1
826 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE31:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT29]], [[TMP104]]
827 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE31]], label [[OMP_ARRAYCPY_DONE32]], label [[OMP_ARRAYCPY_BODY20]]
828 // CHECK2: omp.arraycpy.done32:
829 // CHECK2-NEXT: call void @__kmpc_end_reduce(%struct.ident_t* @[[GLOB3]], i32 [[TMP94]], [8 x i32]* @.gomp_critical_user_.reduction.var)
830 // CHECK2-NEXT: br label [[DOTOMP_REDUCTION_DEFAULT]]
831 // CHECK2: .omp.reduction.default:
832 // CHECK2-NEXT: [[TMP113:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
833 // CHECK2-NEXT: call void @llvm.stackrestore(i8* [[TMP113]])
834 // CHECK2-NEXT: [[TMP114:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
835 // CHECK2-NEXT: [[TMP115:%.*]] = load i32, i32* [[TMP114]], align 4
836 // CHECK2-NEXT: call void @__kmpc_barrier(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 [[TMP115]])
837 // CHECK2-NEXT: ret void
838 //
839 //
840 // CHECK2-LABEL: define {{[^@]+}}@.red_init.
841 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3:[0-9]+]] {
842 // CHECK2-NEXT: entry:
843 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
844 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
845 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
846 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
847 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
848 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
849 // CHECK2-NEXT: store i32 0, i32* [[TMP3]], align 8
850 // CHECK2-NEXT: ret void
851 //
852 //
853 // CHECK2-LABEL: define {{[^@]+}}@.red_comb.
854 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
855 // CHECK2-NEXT: entry:
856 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
857 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
858 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
859 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
860 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
861 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to i32*
862 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
863 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to i32*
864 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP3]], align 8
865 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP5]], align 8
866 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP6]], [[TMP7]]
867 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP3]], align 8
868 // CHECK2-NEXT: ret void
869 //
870 //
871 // CHECK2-LABEL: define {{[^@]+}}@.red_init..1
872 // CHECK2-SAME: (i8* noalias [[TMP0:%.*]], i8* noalias [[TMP1:%.*]]) #[[ATTR3]] {
873 // CHECK2-NEXT: entry:
874 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
875 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
876 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
877 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
878 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
879 // CHECK2-NEXT: [[TMP3:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
880 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr i8, i8* [[TMP2]], i64 [[TMP3]]
881 // CHECK2-NEXT: [[OMP_ARRAYINIT_ISEMPTY:%.*]] = icmp eq i8* [[TMP2]], [[TMP4]]
882 // CHECK2-NEXT: br i1 [[OMP_ARRAYINIT_ISEMPTY]], label [[OMP_ARRAYINIT_DONE:%.*]], label [[OMP_ARRAYINIT_BODY:%.*]]
883 // CHECK2: omp.arrayinit.body:
884 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYINIT_BODY]] ]
885 // CHECK2-NEXT: store i8 0, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
886 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
887 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP4]]
888 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYINIT_DONE]], label [[OMP_ARRAYINIT_BODY]]
889 // CHECK2: omp.arrayinit.done:
890 // CHECK2-NEXT: ret void
891 //
892 //
893 // CHECK2-LABEL: define {{[^@]+}}@.red_comb..2
894 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
895 // CHECK2-NEXT: entry:
896 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
897 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
898 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
899 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
900 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* @{{reduction_size[.].+[.]}}, align 8
901 // CHECK2-NEXT: [[TMP3:%.*]] = load i8*, i8** [[DOTADDR]], align 8
902 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
903 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr i8, i8* [[TMP3]], i64 [[TMP2]]
904 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP3]], [[TMP5]]
905 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
906 // CHECK2: omp.arraycpy.body:
907 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP4]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
908 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP3]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
909 // CHECK2-NEXT: [[TMP6:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
910 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP6]] to i32
911 // CHECK2-NEXT: [[TMP7:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
912 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP7]] to i32
913 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
914 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i32 [[ADD]] to i8
915 // CHECK2-NEXT: store i8 [[CONV3]], i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
916 // CHECK2-NEXT: [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
917 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
918 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
919 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
920 // CHECK2: omp.arraycpy.done4:
921 // CHECK2-NEXT: ret void
922 //
923 //
924 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_privates_map.
925 // CHECK2-SAME: (%struct..kmp_privates.t* noalias [[TMP0:%.*]], i8*** noalias [[TMP1:%.*]]) #[[ATTR6:[0-9]+]] {
926 // CHECK2-NEXT: entry:
927 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca %struct..kmp_privates.t*, align 8
928 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8***, align 8
929 // CHECK2-NEXT: store %struct..kmp_privates.t* [[TMP0]], %struct..kmp_privates.t** [[DOTADDR]], align 8
930 // CHECK2-NEXT: store i8*** [[TMP1]], i8**** [[DOTADDR1]], align 8
931 // CHECK2-NEXT: [[TMP2:%.*]] = load %struct..kmp_privates.t*, %struct..kmp_privates.t** [[DOTADDR]], align 8
932 // CHECK2-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT__KMP_PRIVATES_T:%.*]], %struct..kmp_privates.t* [[TMP2]], i32 0, i32 0
933 // CHECK2-NEXT: [[TMP4:%.*]] = load i8***, i8**** [[DOTADDR1]], align 8
934 // CHECK2-NEXT: store i8** [[TMP3]], i8*** [[TMP4]], align 8
935 // CHECK2-NEXT: ret void
936 //
937 //
938 // CHECK2-LABEL: define {{[^@]+}}@.omp_task_entry.
939 // CHECK2-SAME: (i32 [[TMP0:%.*]], %struct.kmp_task_t_with_privates* noalias [[TMP1:%.*]]) #[[ATTR3]] {
940 // CHECK2-NEXT: entry:
941 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR_I:%.*]] = alloca i32, align 4
942 // CHECK2-NEXT: [[DOTPART_ID__ADDR_I:%.*]] = alloca i32*, align 8
943 // CHECK2-NEXT: [[DOTPRIVATES__ADDR_I:%.*]] = alloca i8*, align 8
944 // CHECK2-NEXT: [[DOTCOPY_FN__ADDR_I:%.*]] = alloca void (i8*, ...)*, align 8
945 // CHECK2-NEXT: [[DOTTASK_T__ADDR_I:%.*]] = alloca i8*, align 8
946 // CHECK2-NEXT: [[__CONTEXT_ADDR_I:%.*]] = alloca %struct.anon*, align 8
947 // CHECK2-NEXT: [[DOTFIRSTPRIV_PTR_ADDR_I:%.*]] = alloca i8**, align 8
948 // CHECK2-NEXT: [[TMP_I:%.*]] = alloca i8**, align 8
949 // CHECK2-NEXT: [[TMP4_I:%.*]] = alloca i8*, align 8
950 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i32, align 4
951 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca %struct.kmp_task_t_with_privates*, align 8
952 // CHECK2-NEXT: store i32 [[TMP0]], i32* [[DOTADDR]], align 4
953 // CHECK2-NEXT: store %struct.kmp_task_t_with_privates* [[TMP1]], %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
954 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTADDR]], align 4
955 // CHECK2-NEXT: [[TMP3:%.*]] = load %struct.kmp_task_t_with_privates*, %struct.kmp_task_t_with_privates** [[DOTADDR1]], align 8
956 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES:%.*]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 0
957 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T:%.*]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 2
958 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T]], %struct.kmp_task_t* [[TMP4]], i32 0, i32 0
959 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
960 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to %struct.anon*
961 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_T_WITH_PRIVATES]], %struct.kmp_task_t_with_privates* [[TMP3]], i32 0, i32 1
962 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast %struct..kmp_privates.t* [[TMP9]] to i8*
963 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast %struct.kmp_task_t_with_privates* [[TMP3]] to i8*
964 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META3:![0-9]+]])
965 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META6:![0-9]+]])
966 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META8:![0-9]+]])
967 // CHECK2-NEXT: call void @llvm.experimental.noalias.scope.decl(metadata [[META10:![0-9]+]])
968 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
969 // CHECK2-NEXT: store i32* [[TMP5]], i32** [[DOTPART_ID__ADDR_I]], align 8, !noalias !12
970 // CHECK2-NEXT: store i8* [[TMP10]], i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
971 // 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
972 // CHECK2-NEXT: store i8* [[TMP11]], i8** [[DOTTASK_T__ADDR_I]], align 8, !noalias !12
973 // CHECK2-NEXT: store %struct.anon* [[TMP8]], %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
974 // CHECK2-NEXT: [[TMP12:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR_I]], align 8, !noalias !12
975 // CHECK2-NEXT: [[TMP13:%.*]] = load void (i8*, ...)*, void (i8*, ...)** [[DOTCOPY_FN__ADDR_I]], align 8, !noalias !12
976 // CHECK2-NEXT: [[TMP14:%.*]] = load i8*, i8** [[DOTPRIVATES__ADDR_I]], align 8, !noalias !12
977 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast void (i8*, ...)* [[TMP13]] to void (i8*, i8***)*
978 // CHECK2-NEXT: call void [[TMP15]](i8* [[TMP14]], i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]]) #[[ATTR5:[0-9]+]]
979 // CHECK2-NEXT: [[TMP16:%.*]] = load i8**, i8*** [[DOTFIRSTPRIV_PTR_ADDR_I]], align 8, !noalias !12
980 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP12]], i32 0, i32 1
981 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[TMP17]], align 8
982 // CHECK2-NEXT: [[TMP19:%.*]] = load i8*, i8** [[TMP16]], align 8
983 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTGLOBAL_TID__ADDR_I]], align 4, !noalias !12
984 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i32* [[TMP18]] to i8*
985 // CHECK2-NEXT: [[TMP22:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP19]], i8* [[TMP21]]) #[[ATTR5]]
986 // CHECK2-NEXT: [[CONV_I:%.*]] = bitcast i8* [[TMP22]] to i32*
987 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
988 // CHECK2-NEXT: [[TMP24:%.*]] = load i8**, i8*** [[TMP23]], align 8
989 // CHECK2-NEXT: [[TMP25:%.*]] = load i8*, i8** [[TMP24]], align 8
990 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 1
991 // CHECK2-NEXT: [[TMP27:%.*]] = load i32*, i32** [[TMP26]], align 8
992 // CHECK2-NEXT: [[TMP28:%.*]] = load i32, i32* [[TMP27]], align 4
993 // CHECK2-NEXT: [[TMP29:%.*]] = sext i32 [[TMP28]] to i64
994 // CHECK2-NEXT: [[LB_ADD_LEN_I:%.*]] = add nsw i64 -1, [[TMP29]]
995 // CHECK2-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
996 // CHECK2-NEXT: [[TMP31:%.*]] = load i8**, i8*** [[TMP30]], align 8
997 // CHECK2-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i8*, i8** [[TMP31]], i64 9
998 // CHECK2-NEXT: [[TMP32:%.*]] = load i8*, i8** [[ARRAYIDX2_I]], align 8
999 // CHECK2-NEXT: [[ARRAYIDX3_I:%.*]] = getelementptr inbounds i8, i8* [[TMP32]], i64 [[LB_ADD_LEN_I]]
1000 // CHECK2-NEXT: [[TMP33:%.*]] = ptrtoint i8* [[ARRAYIDX3_I]] to i64
1001 // CHECK2-NEXT: [[TMP34:%.*]] = ptrtoint i8* [[TMP25]] to i64
1002 // CHECK2-NEXT: [[TMP35:%.*]] = sub i64 [[TMP33]], [[TMP34]]
1003 // CHECK2-NEXT: [[TMP36:%.*]] = sdiv exact i64 [[TMP35]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1004 // CHECK2-NEXT: [[TMP37:%.*]] = add nuw i64 [[TMP36]], 1
1005 // CHECK2-NEXT: [[TMP38:%.*]] = mul nuw i64 [[TMP37]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1006 // CHECK2-NEXT: store i64 [[TMP37]], i64* @{{reduction_size[.].+[.]}}, align 8
1007 // CHECK2-NEXT: [[TMP39:%.*]] = load i8*, i8** [[TMP16]], align 8
1008 // CHECK2-NEXT: [[TMP40:%.*]] = call i8* @__kmpc_task_reduction_get_th_data(i32 [[TMP20]], i8* [[TMP39]], i8* [[TMP25]]) #[[ATTR5]]
1009 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[TMP12]], i32 0, i32 2
1010 // CHECK2-NEXT: [[TMP42:%.*]] = load i8**, i8*** [[TMP41]], align 8
1011 // CHECK2-NEXT: [[TMP43:%.*]] = load i8*, i8** [[TMP42]], align 8
1012 // CHECK2-NEXT: [[TMP44:%.*]] = ptrtoint i8* [[TMP43]] to i64
1013 // CHECK2-NEXT: [[TMP45:%.*]] = ptrtoint i8* [[TMP25]] to i64
1014 // CHECK2-NEXT: [[TMP46:%.*]] = sub i64 [[TMP44]], [[TMP45]]
1015 // CHECK2-NEXT: [[TMP47:%.*]] = sdiv exact i64 [[TMP46]], ptrtoint (i8* getelementptr (i8, i8* null, i32 1) to i64)
1016 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr i8, i8* [[TMP40]], i64 [[TMP47]]
1017 // CHECK2-NEXT: store i8** [[TMP4_I]], i8*** [[TMP_I]], align 8, !noalias !12
1018 // CHECK2-NEXT: store i8* [[TMP48]], i8** [[TMP4_I]], align 8, !noalias !12
1019 // CHECK2-NEXT: ret i32 0
1020 //
1021 //
1022 // CHECK2-LABEL: define {{[^@]+}}@.omp.reduction.reduction_func
1023 // CHECK2-SAME: (i8* [[TMP0:%.*]], i8* [[TMP1:%.*]]) #[[ATTR3]] {
1024 // CHECK2-NEXT: entry:
1025 // CHECK2-NEXT: [[DOTADDR:%.*]] = alloca i8*, align 8
1026 // CHECK2-NEXT: [[DOTADDR1:%.*]] = alloca i8*, align 8
1027 // CHECK2-NEXT: store i8* [[TMP0]], i8** [[DOTADDR]], align 8
1028 // CHECK2-NEXT: store i8* [[TMP1]], i8** [[DOTADDR1]], align 8
1029 // CHECK2-NEXT: [[TMP2:%.*]] = load i8*, i8** [[DOTADDR]], align 8
1030 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8* [[TMP2]] to [3 x i8*]*
1031 // CHECK2-NEXT: [[TMP4:%.*]] = load i8*, i8** [[DOTADDR1]], align 8
1032 // CHECK2-NEXT: [[TMP5:%.*]] = bitcast i8* [[TMP4]] to [3 x i8*]*
1033 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 0
1034 // CHECK2-NEXT: [[TMP7:%.*]] = load i8*, i8** [[TMP6]], align 8
1035 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to i32*
1036 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 0
1037 // CHECK2-NEXT: [[TMP10:%.*]] = load i8*, i8** [[TMP9]], align 8
1038 // CHECK2-NEXT: [[TMP11:%.*]] = bitcast i8* [[TMP10]] to i32*
1039 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP5]], i64 0, i64 1
1040 // CHECK2-NEXT: [[TMP13:%.*]] = load i8*, i8** [[TMP12]], align 8
1041 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 1
1042 // CHECK2-NEXT: [[TMP15:%.*]] = load i8*, i8** [[TMP14]], align 8
1043 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[TMP3]], i64 0, i64 2
1044 // CHECK2-NEXT: [[TMP17:%.*]] = load i8*, i8** [[TMP16]], align 8
1045 // CHECK2-NEXT: [[TMP18:%.*]] = ptrtoint i8* [[TMP17]] to i64
1046 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP11]], align 4
1047 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP8]], align 4
1048 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1049 // CHECK2-NEXT: store i32 [[ADD]], i32* [[TMP11]], align 4
1050 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr i8, i8* [[TMP15]], i64 [[TMP18]]
1051 // CHECK2-NEXT: [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq i8* [[TMP15]], [[TMP21]]
1052 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1053 // CHECK2: omp.arraycpy.body:
1054 // CHECK2-NEXT: [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi i8* [ [[TMP13]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1055 // CHECK2-NEXT: [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi i8* [ [[TMP15]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1056 // CHECK2-NEXT: [[TMP22:%.*]] = load i8, i8* [[OMP_ARRAYCPY_DESTELEMENTPAST]], align 1
1057 // CHECK2-NEXT: [[CONV:%.*]] = sext i8 [[TMP22]] to i32
1058 // CHECK2-NEXT: [[TMP23:%.*]] = load i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], align 1
1059 // CHECK2-NEXT: [[CONV2:%.*]] = sext i8 [[TMP23]] to i32
1060 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[CONV]], [[CONV2]]
1061 // CHECK2-NEXT: [[CONV4:%.*]] = trunc i32 [[ADD3]] to i8
1062 // CHECK2-NEXT: store i8 [[CONV4]], 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_SRC_ELEMENT]] = getelementptr i8, i8* [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1065 // CHECK2-NEXT: [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq i8* [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP21]]
1066 // CHECK2-NEXT: br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1067 // CHECK2: omp.arraycpy.done5:
1068 // CHECK2-NEXT: ret void
1069 //
1070